Transcript Air Quality Forecasting in the Philadelphia

Air Quality Forecasting in the
Philadelphia Metropolitan Area
Bill Ryan
Department of Meteorology
Penn State
[email protected]
National Weather Service, Mount Holly, April 28, 2010
http://www.meteo.psu.edu/~wfryan/mount-holly-2010.pptx
What Do We Forecast?
• Ozone (O3) – since 1996
– Photochemical pollutant, requires hydrocarbons,
oxides of nitrogen (NOx) and UV sunlight.
– Season: April-October, peak in June-August.
• Fine Particles (PM2.5) – since 2003
– Many sources, primarily products of combustion,
many formation mechanisms.
– Year round pollutant, peaks in summer season
(sulfate) with a secondary peak in deep winter.
Why Do We Forecast?
• O3
– At high concentrations, causes “sunburn” to lungs
– Most affected:
• Children: They process a lot of air, affect growth of lungs
and therefore has lifelong impact on lung capacity.
• Elderly and anyone with a compromised respiratory system.
Ozone amplifies response to other allergens.
• PM2.5
– Reduces lung function similar to O3 but particles can
also cross the boundary from the lungs into the
bloodstream and cause cardio-pulmonary problems.
Who Uses Our Forecasts?
• Forecasts are directed to two purposes:
– Protect public health
• Issue watches and warnings well in advance so citizens can
plan their activities.
– Reduce emissions of O3 and PM2.5 “precursors”:
• “Air Quality Action Day”
• Voluntary program of large employers to reduce emissions
across the region on forecast bad air days.
• The lag time required between warnings and
action is why forecasts are issued ~ 2-3 pm daily
and valid the following day (12-36 h forecast).
Where Do We Forecast?
O3 monitors shown in black
triangles.
PM2.5 monitors generally
in same locations but about
half the density.
Air Quality Alerts
Beginning last summer, Air
Quality Alerts were posted to the
NWS forecast page.
The Philadelphia metropolitan
area, as far as the EPA is
concerned, includes portions
of three states.
This poses a challenge for forecast
coordination as the states are
ultimately responsible for meeting
clean air standards.
Warning are coordinated between
neighboring states (PA, NJ, DE).
What Are the Forecast Metrics (Official)?
• Units of Measure
– O3: parts per billion by volume (ppbv)
– PM2.5: micrograms per cubic meter (µgm-3)
• Warning Criteria
– Pegged to National Ambient Air Quality Standard
(NAAQS), standard re-evaluated every 5 years.
– O3: 8-hour running average ≥ 76 ppbv. Currently
under review.
– PM2.5: 24-hour (midnight-midnight) average ≥ 35
µgm-3.
How Are Metrics Reported to Public?
• Air Quality Index (AQI)
– Dimensionless parameter where > 100 is the Warning
Criteria level.
– AQI for any given day is the highest of either O3 or
PM2.5 AQI.
• Color Codes
– Green (“Good”, 0-50 AQI), Yellow (“Moderate”, 51-100
AQI)
– Orange (“Unhealthy for Sensitive Groups”, 101-150
AQI), similar to Watch. Air Quality Alert issued.
– Red (“Unhealthy”, > 151 AQI), similar to Warning.
Climatology of Color Codes for O3 in PHL:
Summer Season (JJA)
1994-2002
2003-2008
5% 1%
4%
12%
29%
19%
41%
27%
28%
34%
Ozone climatology (?) is never static. For example, significant regional scale
controls of power plant NOx emissions were introduced In the 2002-2003 time period.
As result, we have observed a “step down” in O3 to cleaner levels since 2003.
Changes in O3 Climatology:
Frequency of Severe O3 Cases in PHL
30
Days Above Peak Ozone Threshold
> 115 ppbv
25
20
> 105 ppbv
> 96 ppbv Code Red
15
10
5
0
2009 is not blank, there just weren’t any Code Red cases!
Climatology of Color Codes for O3 in PHL
2003-2008
2009
0%
5% 1%
19%
0%
5%
41%
36%
59%
34%
Year to year variations in O3 respond to weather and non-weather factors.
For example, 2009 was cool and wet, but the recession
further reduced O3 precursor emissions of hydrocarbons and NOx.
PHL PM2.5 Concentrations Remarkably Low in 2009:
Function of Lower Energy Usage
Number of Days Above PM2.5
Concentration Threshold (microg/m3)
25
23
20
18
18
17
15
15
12
13
> 35.5
> 30.0
10
6
7
8
5
0
0 0
2004
2005
2006
2007
2008
2009
Note: Data for 2004-2008 uses gravimetric filter monitors (FRM) while 2009
uses 24-average from continuous monitors.
What Are Our Forecast Tools?
• Persistence
– Both O3 and PM2.5, and some of their precursors, have long
lifetimes in the troposphere (on order of days to week)
– As a result, local air quality is a combination of local and
transported pollutants.
– What is the direction of transport and the concentrations
of pollutants being transported?
• Weather Parameters
– O3 is associated with high temperature, few clouds.
– PM2.5 has many sources and formation routes, therefore,
no simple weather associations. In summer, humidity
increases sulfate load.
What Are Our Forecast Tools?
• Statistical Models
– Useful for O3, cost-effective though limited
because need long training period.
– Not useful for PM2.5.
• Numerical Models
– NOAA-EPA Operational O3 model in use since 2005
(http://weather.gov.aq). Several other models,
see Appendix A.
– PM2.5 model in developmental stage at NOAA.
Other PM2.5 models are available but skill limited.
What is Our Forecast Skill?
• For O3:
– Median Absolute Error: 6.2 ppbv, or ± 10%
– Bias: +1.7 ppbv
• For PM2.5:
– Median Absolute Error:
• Summer Season: 4.0 µgm-3, or ± 20%
• Overall: 3.5 µgm-3
– Bias
• Summer Season: + 0.4 µgm-3
• Overall: : + 0.4 µgm-3
Forecast Skill for Air Quality Alerts
(May-September, 2004-2009)
• Standard Measures
– Hit Rate: 74%
– False Alarm: 31%
– Accuracy: 87%
• Skill Measures:
– Threat Score: 0.56, Range: [0,1]
– Heidke Score: 0.63, Range: [-1,1]
• Skill Score with Reference to Persistence:
– Improvements of 40-53% depending on measures
selected for comparison.
Forecast Skill for Air Quality Alerts
(May-September, 2004-2009)
• Forecast skill primarily driven by good O3
forecasts. PM2.5 forecasts are less skillful:
– Hit Rate for PM2.5 is only 43%, but 82% of the
“missed” PM2.5 cases were covered by an Air
Quality Alert for O3 already in place.
– Frequency of Code Orange PM2.5 cases in warm
season is 2.3 times less than Code Orange O3.
– Worth noting: Most bad PM2.5 days are also bad
O3 days (84%), but bad O3 days are not usually bad
PM2.5 days (23%).
What Does a Poor Air Quality Day Look Like?
• O3
– Full sun, long day length (high SZA)
– Light winds, transport aloft (0.5-2 km) from west;
pollutant concentrations in the residual layer is critical
– Limited vertical mixing
• PM2.5
–
–
–
–
Light winds
Transport aloft from west
Strong morning inversion and limited mixing later
High humidity
Poor air quality events are linked to
the synoptic cycle. The majority
of bad air days occur in multi-day
episodes (2-5 days).
In the standard summer season
pollution episode, an upper air
ridge, with its axis over or west of
the mid-Atlantic, is in place.
A cold front passes bringing clean
air, then, as surface high pressure
migrates slowly from the Midwest,
it becomes modified (dirty).
H
H
Eventually the continental high pressure
center stalls and links up with the semipermanent Bermuda High circulation.
This is a common summer season PM2.5
and O3 episode pattern. High pressure
overhead leads to clear skies, and light
winds.
This is not, by itself, enough to lead to
a bad air day. Hot weather and light
winds are necessary, but not
sufficient, for poor air quality.
H
H
Example: Surface Analysis During a Poor Air
Quality Event
An Appalachian Lee Trough is commonly
associated with poor air quality cases.
Other Necessary Ingredients
• In addition to sunny skies and light winds,
need:
– Limited Vertical Mixing
• Some kind of trapping inversion. In the summer
season, surface-based inversions break early in the day.
A secondary inversion aloft (950-850 mb) or strong
warm air advection is often necessary.
– Transport of Pollutants into Region
• The Ohio River Valley is the largest source of power
plant emissions of NOx (for O3) and sulfate (for PM2.5).
Westerly Winds in Residual Layer Bring
Pollutants to Our Forecast Area
Average 850 mb Winds
July 12-17, 1997
Vector averaged winds at
850 mb during an unhealthy
O3 episode in 1997.
Midwest and Ohio River
Valley are source of high
emissions of oxides of
nitrogen (forms O3 and
PM2.5) and sulfates
(forms PM2.5).
Emissions controls enacted
in 2003 in that region have
reduced O3 significantly.
Back Trajectory Forecast Models Help Identify
Source Regions for Tomorrow’s Air Quality
HYSPLIT Back Trajectory
Colored lines show the forecast path
of air parcels reaching PHL at 1200 UTC
on July 16 at three levels above
the ground (500, 1000, 1500 m).
NOAA ARL HYSPLIT Model
http://www.air.noaa.gov/ready
The path is 24 hours in duration
with dots giving position at
6 hour intervals.
The bottom panel shows the forecast
vertical motion of the parcels.
Trajectories are coupled with real time
AQ data to estimate upwind contribution
to tomorrow’s air quality.
When Good Forecasts Go Bad
• Convection
– Local Convection: Timing is critical for O3. Convection
after ~ 6 pm is usually too late to clean the air.
– “Near-Local” Convection: Downdrafts within outflow
boundaries contain very clean air.
– Regional Convection: Reduces transported pollutants
and increases high clouds; e.g., evening/overnight
MCS upwind.
• Stalled/Reversing Frontal Boundaries
• Sea/Bay Breeze Re-circulations
Effect of Outflow Boundaries on O3
Fair Hill, MD
The effect of thunderstorms on local O3 can be
GOES Visible
remarkable even at its periphery.
1902 UTC
June 26, 1998
Hourly Ozone Concentrations
Fair Hill, MD
June 25-26, 1998
Regional Scale Convection and Local O3
24-h back trajectories at 500,
1000 and 1500 m
Aug 4
Aug 5
O3 Concentrations, 2 pm, August 4
August 5th was forecast to be sunny and very warm - conducive to O3 formation.
Back trajectories are from climatologically “dirty” location. But, early afternoon O3
concentrations on August 4th along the path of trajectory are clean!
Low Upwind O3 Concentrations Due to Organized
Thunderstorm System (MCS) on August 4
GOES IR4
1515 UTC, August 4
MCS moved through OH on the
morning of the previous day
Result: While O3 was high in PHL the
following day, no locations reached the
8-hour average Code Orange threshold.
Summary (1 of 2)
• Air quality forecasts (O3 and PM2.5) are issued
daily for the Philadelphia metropolitan area,
and surrounding states.
• Forecasts are issued ~ 2pm in a color code
format with Alerts issued for Code Orange or
higher.
• The peak season for poor air quality is JuneAugust although PM2.5 can reach Code Orange
levels year round.
Summary (2 of 2)
• Poor air quality typically occurs in multi-day
episodes linked to the synoptic cycle.
• Warm temperatures, light winds, sunny skies,
limited vertical mixing, with westerly winds in the
residual layer, are associated with poor air quality
episodes.
• Forecasts are reasonably accurate but become
more uncertain in the presence of convection
(local and regional) as well as stalled frontal
boundaries and sea breeze fronts.
Acknowledgements
• Air quality forecasting in the Philadelphia area
is funded by the Delaware Valley Regional
Planning Commission (DVRPC) in association
with the states of Pennsylvania, Delaware and
New Jersey.
• Special thanks to Sean Greene of DVRPC as
well as my forecasting colleagues in the midAtlantic region.
Appendix A: Useful Web Sites
• Current Conditions
– EPA AirNow
• Forecast Models
– NOAA-EPA O3 Model
– NC DENR Model (off line at this time, back in May?)
– Environment Canada
• Forecasts and Blogs
–
–
–
–
Philadelphia Metropolitan Area Air Quality Forecast
Mid-Atlantic Medium Range Air Quality Outlook
The Smog Blog
Twitter: @aqforecast