Docstoc

forecasting

Document Sample
forecasting Powered By Docstoc
					  PM2.5 Forecasting Method
Development and Operations for
     Salt Lake City, Utah
                   Presented by
        Timothy S. Dye Dianne S. Miller
    Craig B. Anderson Clinton P. MacDonald
    Charley A. Knoderer Beverly S. Thompson
            Sonoma Technology, Inc.
                  Petaluma, CA
                 (707) 665-9900
             www.sonomatech.com
             tim@sonomatech.com

Presented at EPA’s National Air Quality Conference:
             Mapping and Forecasting
                San Francisco, CA
                February 4-6, 2002
                                                      901491-2146
       Objective and Outline
• Objective: Develop PM2.5 forecasting
  methods and forecast PM2.5 for Salt Lake City
  area during winter 2002
• Outline:
  – Data analysis
  – Forecasting method (tool) development
  – Operational forecasting
  – Initial results
  – Data issues
  – Future plans
                                                  2
                     Data Analysis
•   Description of data sources
     – PM2.5 and PM10 data (AIRS database)
         • Winter months (Oct. – Mar.)
         • 5 years of PM10 data (1996 – 2001)
         • 2 to 4 years of PM2.5 data (1998 – 2001)
         • 14 sites with PM10 data, 16 sites with PM2.5 data
         • Used only the 24-hr filter data
         • Continuous PM2.5 and PM10 data from three TEOM monitors
           (1999 – 2001)
     – Meteorological data (NOAA)
        • 5 years of hourly surface observations and upper-air
          soundings from Salt Lake City airport
        • Daily Weather Maps
•   Data processing
         • Performed quality control checks and time/units standardization
         • Used an MS Access database to store raw and computed data
           values
                                                                             3
Data Analysis




                4
                   Data Analysis
• Developed a climatology: Examined the
  frequency and characteristics of PM episodes in
  Salt Lake City
• Examined the following based days (based on PM2.5) for Salt Lake City
                                           AQI
                                on anMonthly Distribution of AQI
                                 Average


  computed from PM2.5:       35
                                                                     Good
      – Yearly trend                                    30           Moderate




                                       Number of days
                                                        25           Unhealthy for SG

      – Monthly frequency                               20
                                                                     Unhealthy


      – Day of week frequency                           15

                                                        10
• Overview of results                                    5

                                                         0
      Higher PM2.5 tends to occur                            Oct   Nov   Dec      Jan   Feb   Mar
      – on Thursday – Sunday                         Average monthly distribution of AQI
                                                 categories (based on PM2.5) for Salt Lake City
      – in December and January
                                                                                                    5
                               Data Analysis
                                                   Good
                    20                             Moderate
                                                   Unhealthy for SG
                    18                             Unhealthy

                    16
   Number of days




                    14
                    12
                    10
                     8
                     6
                     4
                     2
                     0
                         Sun   Mon   Tue   Wed   Thu    Fri       Sat


Day-of-week frequency for AQI categories (based on PM2.5)
         in the Salt Lake City region (1998-2001)

                                                                        6
Forecasting Method Development
• Analyzed PM2.5 episodes for general
  relationships in weather conditions and
  identified several important variables
  – 500-mb pattern (ridge or trough overhead)
  – Strength and duration of temperature inversion
  – Surface wind speed and direction
• Developed forecast guidelines based on these
  variables (subjective)
• Developed regression equations and
  Classification and Regression Tree (objective)

                                                     7
                 Forecast Guidelines
 Forecast PM2.5 ranges and associated meteorological conditions
    0-15 ug/m3           15-25 ug/m3           25-40 ug/m3           >40 ug/m3
Strong trough         Weak ridge or        Moderate ridge       Strong ridge
                      trough
No inversion          Weak inversion       Moderate inversion   Strong inversion

                      Light to moderate    Light winds          Light, decoupled
                      winds                                     winds
                                           Moderate surface     Strong surface high
                                           high
                                           Weak daytime         Weak daytime
                                           heating              heating
                                                                Morning haze or fog

If holiday, add       If holiday, add      If holiday, add      If holiday, add
0 ug/m3 to forecast   10 ug/m3 to forecast 20 ug/m3 to forecast 20 ug/m3 to forecast

                                                                                       8
Forecasting Method Development
Objective Tools
  • Produced scatter plots of PM2.5 and meteorological data

  • Ran CART and linear regression to develop equations

  • Resulting regression equation:

  Next Day PM2.5 = 53.4 + 3.4*Holiday – 0.2*Precip – 0.5*WSday
    + 1.0*(700T12Z – Tmin) + 0.8*(700T00Z – Tmax) +
    0.2*700Td00Z – 0.3*850WS00Z – 0.3*Tmax

  • Checked physical relationship between each variable
    and PM2.5

  • Tested both techniques operationally during two weeks
    of forecasting

                                                                 9
  Data Acquisition and Dissemination
              STI Weather Center
 Monitoring
   Sites
 AQ and Met                               E-mail
                                        forecast to
                                         Utah DEQ



               - Data collection
               - Formatting
               - Quality control              STI
               - 24-hr average calcs.    SmogWatch
               - Conversion factor      web page for
Phone                                   Salt Lake City
               - AQI calculations
               - Forecast creation
Utah DEQ


                                                         10
            Daily Forecasting
0900 - 1000 MST
   – Verify that current data is being received
1100 MST
   – Review previous day’s observations and forecast
   – Review current day’s AQ data, weather forecast data,
     and NWS discussions
   – Run regression equation
   – Run CART
   – Create and review current- and next-day forecasts
1200 MST
   – E-mail forecast to UDEQ
Afternoon
   – Monitor conditions for any dynamic changes


                                                            11
         Forecast Dissemination
                            Sample Daily E-mail
Daily PM2.5 Forecast for Salt Lake City, UT

Today's Date: January 25, 2002

Yesterday's regional maximum AQI:       49 (15 ug/m3) - Good
Today's forecasted regional maximum AQI: 60 (20 ug/m 3) - Moderate
Tomorrow's forecasted regional maximum AQI: 60 (20 ug/m 3) - Moderate

Discussion:
The upper-level low pressure system that was expected to move towards Utah today
has slowed some resulting in a stronger inversion than originally anticipated. The
models keep the inversion intact today despite increased winds aloft. This will result in
Moderate PM2.5 values for today. The upper-level low pressure system moves towards
Utah tomorrow increasing winds aloft; however, the inversion is forecasted to remain
intact through Sunday resulting in Moderate PM2.5 levels.

Forecaster: Charley Knoderer

Sonoma Technology, Inc.
(707) 665-9900


                                                                                            12
Forecast Dissemination




  Salt Lake City SmogWatch Web Page
                                      13
               Initial Results
Slight tendency to overpredict the next-day forecast




                                                       14
               Data Issues
• Continuous PM2.5 values from the TEOMs
  can be significantly less than the FRM filter
  values. Difference varies, based on several
  meteorological parameters.

• Continuous PM2.5 monitoring network is
  limited - only one site in each air basin.

• How to communicate hourly AQI when
  “standard” is based on a 24-hour average?

                                                  15
 Data Issues: TEOM vs. FRM
• Collocated TEOM (continuous) and FRM (filter)
• TEOM PM2.5 underestimates the FRM
                         80


                         70
                                                                           Unhealthy        PM2.5 Continuous vs. Filter
                         60
                                                                                            Intercomparison for the
                                                                                            Lindon site
 PM 2.5 FILTER (ug/m3)




                         50
                                                                                            (24-hr average data from
                                                                                  USG
                         40                                                                 December to March of
                                                                                            1999, 2000, and 2001)
                         30
                                                         Continuous data
                                                         underestimates
                         20                               AQI categories
                                                                            Moderate

                         10
                                                                           R2 = 0.63
                         0
                              0   10   20   30      40      50       60      70        80
                                            PM 2.5 TEOM (ug/m3)

                                                                                                                          16
  Data Issues: TEOM vs. FRM
• Adjusted TEOM data using a site specific regression equation that
  incorporated temperature and humidity
• Reduces underestimation by TEOM
                          80


                          70
                                                                                Unhealthy
                                                                                                 Adjusted PM2.5 Continuous
                          60
                                                                                                 vs. Filter Intercomparison
   PM2.5 Filter (ug/m3)




                          50
                                                                                                 for the Lindon site

                                                                                       USG       (24-hr average data from
                          40
                                                                                                 December to March of
                          30                                                                     1999, 2000, and 2001)

                                                                                                 Tc = –1.98 + 1.08 TEOM
                          20
                                                                                 Moderate
                          10                                                                        – 0.67 Avg. Temp.
                                                                                R2 = 0.84           + 0.08 Avg. RH
                          0
                               0   10   20      30       40       50       60     70        80
                                             Adjusted PM2.5 TEOM (ug/m3)

                                                                                                                              17
               Next Steps
• Implement relative humidity and temperature
  conversion factor to adjust continuous TEOM
  data to better match FRM filter data.
• Continue forecasting until March 1
• Modify and adjust forecasting tools as
  “new truths” are learned
• Continue to evaluate forecast performance
• Acknowledgments
  – Utah Department of Environmental Quality
  – U.S. Environmental Protection Agency


                                                18

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:7
posted:7/30/2012
language:English
pages:18