Title: Developing and Operating Air Quality Forecasting Programs for Five Cities
1Developing and Operating Air Quality Forecasting
Programs for Five Cities
- Prepared by
- Clinton P. MacDonald
- Charley A. Knoderer
- Timothy S. Dye
- Beverly S. Thompson
- Dianne S. Miller
- Craig B. Anderson
- Sonoma Technology, Inc.
- Petaluma, CA
- EPA National Air Quality Conference
- Forecasting and Public Outreach
- February 4-6, 2002
- San Francisco, CA
901201-2148
2Outline
- Overview of the programs
- Forecasting tools and process
- Daily operations
- Meteorological and air quality processes
- Similarities between regions
- Local issues and processes
- Verification
- All programs
- Tool comparison
- Observations
3Overview of Programs
- Five cities - three sponsors
- 1. Sacramento, CA
- 6th year forecasting
- 2. Columbus, OH
- Nashville and Memphis, TN
- 1st year forecasting
- 3. Minneapolis, Minnesota
- 1st year forecasting
4Overview of Programs
- Issued forecasts for current- and next-days
regional 8-hr peak additional 5-day outlook in
Sacramento - Forecasts issued by late morning/early afternoon
- Operated 7 days per week
- Season
- May 1 October 31 for Sacramento
- May 1 September 30 for other cities
5Overview of Programs
- Program activities
- Collected and analyzed historical data to
understand local air quality and meteorology - Developed forecasting tools
- Developed infrastructure
- Data acquisition
- Forecasting procedures
- Forecast dissemination methods
- Interacted with clients and media
- Verified and analyzed forecasts
6Forecasting Tools
- Qualitative and quantitative tools that aid
forecasters in predicting air quality
1 Regression equations were developed and used
at the end of the ozone season. 2 Photochemical
models were developed and used at the end of the
ozone season.
7Forecasting Tool - Criteria
- Minneapolis Why Criteria?
- Difficult to develop accurate objective methods
for this region - Regression equations had poor correlations
- Best r2 was 0.57
- r2 of at least 0.65 is desired to have any
confidence in the equation - Two causes for poor correlations and inability to
predict high ozone concentrations - Equation does not capture ozone transport
- Too few high ozone days in the developmental
dataset to be reliable
8Forecasting Tool - Criteria
- Guidelines for ozone 80 ppb Transport Case
- Moderate (6 m/s) southerly winds overnight
- Maximum temperature 80ºF
- Average daytime (9 a.m. to 6 p.m.) relative
humidity - At least Moderate AQI 8-hr ozone levels upstream
on previous day - 500-mb ridge axis east of Minneapolis
9Forecasting Tool - CART
- Sacramento, CA
- A decision tree was developed in 1999
- Easy to use with decent accuracy
10Forecasting Process
- Used a variety of tools to reach a final forecast
11Daily Operations
- Timeline of activities
- Quick-cast to decide where to spend effort each
day - Staffing to plan for simultaneous high ozone days
in multiple regions - Disseminate forecast to districts through
multiple methods to ensure delivery - Direct e-mail, fax, phone
- Indirect SmogWatch and AIRNow web pages
12Meteorological and Air Quality Processes -
Similarities
- Ridge
- Causes sinking motion
- Inversion that traps pollutants and limits
vertical mixing - Generally light surface winds
- Warm temperatures
- Mostly clear skies
13Meteorological and Air Quality Processes - Local
Issues
- Many large-scale similarities, but smaller scale
meteorology and air quality must be understood to
produce accurate forecasts - How large-scale weather influences local weather
- Local carryover and transport
- Local flow patterns
- Importance of thunderstorms and clouds
- Local emissions patterns
- Monitor locations and characteristics
- Event duration
- Ramp-up time
- Using several forecasting tools helps better
capture the influence of these phenomena on air
quality
14Local Processes
- Sacramento, CA
- Hot spots of ozone due
- to smoke from forest fires
- Columbus, OH
- Clouds and weak fronts are critical
- Difficult to predict the exact position of these
features 12 to 36 hrs in advance
15Local Processes
- Memphis, TN
- Monitoring network is limited
- Memphis peak can change with slight wind
direction shift
8-hour maximum of .091 ppm at Crittenden and 0.70
ppm at Frayser
8-hour maximum of 0.114 ppm at Frayser
16Local Processes
- Minneapolis, MN Long-range transport is
important
General 24-hour air parcel paths for a June 2001
ozone episode based on trajectories. Created
using NOAAs Hysplit trajectory model.
17Verification - All Programs
- Verify next-day (24-hr) forecast
- Compare observed and forecasted categories
- Focus on simple performance measures
- Evaluate two-category forecast
- GoodModerate
- Unhealthy
18Verification - All Programs
- Compute performance measures
- Percent correct (PC) - Percent of forecasts that
correctly predicted the Good-Moderate or
Unhealthy AQI categories - False alarm (FA) - Percent of times a forecast of
Unhealthy did not actually occur (crying wolf). - Probability of detection (POD) - Percent of
actual Unhealthy days correctly predicted.
19Verification - All Programs
- Verification results for all cities
20Verification Tool Comparison
- Two ozone forecasts issued in Nashville
- Objective using regression (TVA)
- Subjective using conceptual/experience (STI)
- Independently forecasted throughout summer 2001
- Compared these two approaches to evaluate their
strengths and limitations
21Verification Tool Comparison
22Verification Tool Comparison
- Compute verification metrics
- Evaluate ability to predict Unhealthy AQI
23Observations
- Regression underpredicts peaks (lower POD)
- Conceptual/experience method overpredicts peaks
(higher bias), but correctly predicts more peak
days (higher POD) - Combined forecast (objective with subjective)
beats both the human forecasts alone and
regression alone!
24Acknowledgments
- Sponsors
- U.S. Environmental Protection Agency
- Sacramento Metropolitan Air Quality Management
District - Lake Michigan Ozone Directors Consortium
- Agencies
- Minnesota Pollution Control Agency
- The Mid-Ohio Regional Planning Commission
- Ohio Environmental Protection Agency
- Tennessee Department of Environment and
Conservation - Metropolitan Nashville and Davidson County Health
Department, Pollution Control Division - Memphis and Shelby County Health Department
- AIRNow program staff
- California Air Resources Board