Aerosol Characterization and the Supporting Information Infrastructures

1
Aerosol Characterization and the Supporting
Information Infrastructures

• R.B. Husar
• Washington University in St. Louis

2
Regional Haze Rule Natural Background by 2064
Natural haze is due to natural windblown dust,
biomass smoke and other natural processes
Man-made haze is due industrial activities AND
man-perturbed smoke and dust emissions A fraction
of the man-perturbed smoke and dust is assigned
to natural by policy decisions

• The goal is to attain natural conditions by 2064
• The baseline is established during 2000-2004
• The first SIP Natural Condition estimate in
  2008
• SIP Natural Condition Revisions every 10 yrs

3
NAAMS National Ambient Air Monitoring Strategy
and NCore
4
FASTNET and DataFed pursues several NAAMS
recommendations

• Insightful Measurements
• Enhanced real-time data delivery to public
• Increase capacity for hazardous air pollutant
  measurements
• Increase in continuous PM measurements
• Support for research grade/technology transfer
  sites
• Auxiliary non-EPA data support
• Multiple pollutant monitoring
• Integration of sources, processes, effects
• Incorporate technological advances
• Information transfer technologies
• Continuous PM monitors
• High sensitivity instruments
• Model-monitor integration

5
FASTNET and DataFed

• FASTNET (Fast Aerosol Sensing Tools for Natural
  Event Tracking) an open communal information
  sharing facility to study aerosol events,
  including detection, tracking and impact on PM
  and haze.
• The main asset of FASTNET is the community of
  data analysts, modelers, managers participating
  in the production of actionable knowledge from
  data and models

The community is supported by a non-intrusive
data integration infrastructure based on Internet
standards (web services) and a set of web-tools
evolving under the federated data system,
DataFed DataFed is supported by its community
and is under the umbrella of the interagency
Earth Science Information Partners, ESIP (NASA,
NOAA and EPA)
6
Aerosol Characterization

• Atmospheric aerosol system has three extra
  dimensions (red), compared to gases (blue)
• Spatial dimensions (X, Y, Z)
• Temporal Dimensions (T)
• Particle size (D)
• Particle Composition ( C )
• Particle Shape (S)
• Bad news The mere characterization of the 7Dim
  aerosol system is a challenge
• Spatially dense network -X, Y, Z(??)
• Continuous monitoring (T)
• Size segregated sampling (D)
• Speciated analysis ( C )
• Shape (??)
• Good news The aerosol system is self-describing.
  
• Once the aerosol is characterized (Speciated
  monitoring) and multidimensional aerosol data are
  organized, (see RPO VIEWS effort), unique
  opportunities exists for extracting information
  about the aerosol system (sources,
  transformations) from the data directly.
• Analysts challenge Deciphering the handwriting
  contained in the data
• Chemical fingerprinting/source apportionment

7
Technical Challenge Characterization

• PM characterization requires many sensors,
  sampling methods and analysis tools
• Each sensor/method covers only a fraction of the
  8-D PM data space.
• Most of the 7-8 Dim PM data space is extrapolated
  from sparse measured data
• Others sensors integrate over time, space,
  chemistry, size etc. .
• Example Satellites, have high spatial resolution
  but integrate over height H, size D, composition
  C, shape, and mixture dimensions these data need
  de-convolution of the integral measures.

• For these sensors, the integral samples need to
  be separated into components

8
Satellite Detection of Fire and Smoke
9
Fire Zones of North America

• FIRE and Norm. Diff. Veg. Index, NDVI
• The Northern zone from Alaska to Newfoundland
  has large fire patches, evidence of large,
  contiguous fires.
• The Northwestern zone (W. Canada, ID, MT, CA)
  is a mixture of large and small fires
• The Southeastern fire zone (TXNCFL) has a
  moderate density of uniformly distributed small
  fires.
• The Mexican zone over low elevation C America
  is the most intense fire zone, sharply separated
  from arid and the lush regions.
• Fires are absent in arid low-vegetation areas
  (yellow).

10
Seasonality of Fire

• Dec, Jan, Feb is generally fire-free except in
  Mexico, and W. Canada
• Mar, Apr, May is the peak fire season in Mexico
  and Cuba fires occur also in Alberta-Manitoba
  and in OK-MO region
• Jun, Jul, Aug is the peak fire season in N.
  Canada, Alaska and the NW US.
• Sep, Oct, Nov is fire over the Northwest and
  the Southeast

11
Seasonal Pattern of Fires over N. America

• The number of ATSR satellite-observed fires peaks
  in warm season
• Fires are random onset and smoke amount is
  unpredictable

12
MISR Seasonal AOT (MISR Team)

• Major smoke emission regions by season

13
Smoke types blue, yellow, white
Quebec Smoke 2002
California Smoke 1999

• Smoke from major fires comes in different colors,
  e.g. blue, yellow.
• The chemical, physical and optical
  characteristics of smokes are not known
• Can the reflectance color be used to classify
  smokes?
• Can column AOT be retrieved for optically thick
  smoke? Multiple scattering, absoption?

14
SeaWiFS, TOMS, Surface Visibility, May 98

• Satellite image of color SeaWiFS data, contours
  of TOMS satellite data (green) and surface
  extinction coefficient, Bext
• The smoke plume extends from Guatemala to Hudson
  May in Canada
• The Bext values indicate that the smoke is
  present at the surface

Surface ozone depressed under smoke
15
Hourly PM10 During the Smoke Event
Hourly PM10 concentration pattern at six eastern
US locations during May 1998.
16
(No Transcript)
17
Fire Pixels from MODIS, June 25-July 6, 2002
Manitoba Sask. Fires Note pixel clusters due
to larger fires
Quebec Fires Note pixel clusters due to larger
fires
SE US Fires Random pixels from small fires

• Several satellite sensor (MODIS, GOES, AVHRR,
  ATSR..) detect the location of most fires -
  DAILY
• These fire pixels can be used as sensor-based
  inputs to regional/global models, e.g. NAAPS
• However, the quantity of smoke emitted from the
  from the fire pixels can not be estimated well
  .
• Hence, real-time model simulation of smoke
  transport is limited by the smoke emission
  estimates

18
MODIS The Fine-Scale Picture The Fires and the
Smoke Transport of Smoke from N. Quebec to SE
Canada and NE US.
MODIS Land Rapid Response System

• 020705MODIS

020706 MODIS
020707 MODIS
19
Haze WebCams
CamNet -Webcam
Hartford, CT
New York, NY
Boston, MA

• 020706 1000 Normal bluish haze

020707 1000 Yellow smoke
20
GOES 8 Animation
July 6 animation low-resolution, high resolution
July 7 animation low resolution, high resolution
21
GOES 8 and ASOS Visibility July 6, 2002 815,
1215, 1615 EST
The largest circles correspond to gt 100 ug/m3
PM2.5
GOES8 20020706_1315 UTC
GOES8 20020706_1715 UTC
GOES8 20020706_2115 UTC
GOES8 20020706_1315
22
Smoke Events Community Websites

• er

23
Smoke Plumes over the Southeast
R 0.68 mm G 0.55 mm B 0.41 mm

• Satellite detection yields the origin and
  location is the shape of smoke plumes

0.41 mm

• The influence of the smoke is to increase the
  reflectance ant short wavelength (0.4 mm)
• At longer wavelength, the aerosol reflectance is
  insignificant.

0.87 mm
24
Cumulative Seasonal PM2.5 Composition

• PM2.5 chemical components were calculated based
  on the CIRA methodology
• In addition, the the organics were (tentatively)
  further separated as Primary Smoke Organics (red)
  and Remainder organics (purple)
• PSO 20(K - 0.15Si 0.02 Na)
• Remainder Org Organics - PSO
• Also, the Unknown mass (white area) is the
  difference between the gravimetrically measured
  and the chemically reconstructed PM2.5.
• The daily chemical composition was aggregated
  over the available IMPROVE data range (1988-99)
  to retain the seasonal structure.
• I order to reduce the noise the daily data were
  smoothed by a 15-day moving average filter.

Shenandoah
25
Peripheral Sites Chemical Mass Balance

• Eastern N. America is surrounded by aerosol
  source regions such as Sahara and Central
  America.
• As a consequence, the PM concentration at the
  edges ranges between 4-15 ug/m3 much of it
  originating outside.
• The chemical composition of the inflow varies by
  location and season.
• At the Everglades, organics, smoke organics and
  LAC dominate over sulfate and fine dust
• Sahara dust, and smoke from Central America and
  W. US/Canada are the main contributions to
  Everglades, FL, and Big Bend, TX.

Voyageurs (scale 0-15 ug/m3)
Acadia
Badlands (scale 0-15 ug/m3)
Big Bend (scale 0-15 ug/m3)
Everglades
26
Peripheral Sites Carbonaceous Mass Balance

• At the northern peripheral sites, Badlands,
  Voyageurs and Acadia, the organics range from 1.5
  to 4 mg/m3
• At Big Bend the organics show a spring peak, with
  a majority of smoke organics. This indicates
  biomass smoke origin.
• At the Everglades, the fall peak is due to
  organics, while smoke organics light absorption
  is present throughout the year.

Voyageurs (scale 0-15 ug/m3)
Badlands (scale 0-15 ug/m3)
Acadia (scale 0-15 ug/m3)
Big Bend (scale 0-15 ug/m3)
Everglades (scale 0-15 ug/m3)
27
Possible Smoke Emission EstimationLocal Smoke
Model with Data Assimilation
Continuous Smoke Emissions
Assimilated Smoke Emission for Available Data
Local Smoke Simulation Model
e..g. MM5 winds, plume model
Assimilated Smoke Pattern
Assimilated Fire Location
Satellite Smoke
Surface Smoke
Fire Location
AOT Aer. Retrieval
Fire Pixel, Field Obs
Visibility, AIRNOW
28
Kansas Agricultural Smoke, April 12, 2003
Organics 35 ug/m3 max
Fire Pixels
PM25 Mass, FRM 65 ug/m3 max
Ag Fires
SeaWiFS, Refl
SeaWiFS, AOT Col
AOT Blue
29
(No Transcript)
30
ASOS Visibility Monitoring System (1200 Sites)

• The Automated Surface Observing System, ASOS
  weather every minute.
• The forward scattering (30-500) visibility sensor
  has a range 17 ft to 30 miles.
• The synoptic visibility data are truncated (lt1/4,
  1/4,..10 miles)
• For smoke and haze events (vis. lt 10 mile)
  truncation not a problem

31
Diurnal Cycle Surface Bext, April 12, 2003
Night
Smoke
00
02
04
06
Day
08
10
12
14
Night
16
18
20
22
High Night Bext
Low Day Bext
32
FASTNET Event Report 040705July4Haze, July 6,
2004July 4, 2004 Aerosol Pulse

• Event Summary by the FASTNET Community
• Please send PPT slides or comments to Erin
  Robinson or Rudy Husar, CAPITA
• Visit the event discussion forum

33
July 4, 2004 Aerosol Pulse

• The US-avg. AIRNOW PM25 shows a 3 hr. spike at
  midnight
• In the (airport) ASOS the July 4 spike is
  conspicuously absent
• Thus, the US spike is due to the urban sites
  affected by smoke

AIRNOW PM25
2000
AIRNOW PM25 US Hourly Average
Pulse
0000
ASOS Bext US Hourly Average
No Pulse
0400
0800
34
Previous work The July 4th Potassium Spike
(Poirot 1998)

• Potassium nitrate is a major component of all
  fireworks (provides the bang!).
• Fine particle K for all IMPROVE data (1988-1997)
  were averaged for each day of year
• The potassium spike on July 5 is 120 ng/m3
  compared to 40-60 during the year
• The corresponding IMPROVE-average daily fine mass
  did not show the spike
• The K spike is clearly something to consider (and
  perhaps screen out) in conducting any analyses
  using K data

35
D U S T Update

• Global Local Dust Over N. America

36
Sahara PM10 Events over Eastern US
July 5, 1992
Much previous work by Prospero, Cahill, Malm,
Scanning the AIRS PM10 and IMPROVE chemical
databases several regional-scale PM10 episodes
over the Gulf Coast (gt 80 ug/m3) that can be
attributed to Sahara.
June 30, 1993
June 21 1997

• The highest July, Eastern US, 90th percentile
  PM10 occurs over the Gulf Coast ( gt 80 ug/m3)
• Sahara dust is the dominant contributor to peak
  July PM10 levels.

37
Origin of Fine Dust Events over the US
Gobi dust in spring Sahara in summer
Fine dust events over the US are mainly from
intercontinental transport
38
FASTNET Event Report 040219TexMexDust
Texas-Mexico Dust Event February 19, 2004
Contributed by the FASNET Community Correspondence
to R Poirot, R Husar
MODIS Rapid Response
39
High Wind Speed Dust Spatially Correspond

• The spatial/temporal correspondence suggests that
  most visibility loss is due to locally suspended
  dust, rather than transported dust
• Alternatively, suspended dust and high winds
  travel forward at the same speed
• Wind speed animation Bext animation. (material
  for model validation?)

40
PM10 gt 10 x PM25During the passage of the dust
cloud over El Paso, the PM10 concentration was
more than 10 times higher than the PM2.5
Schematic

• AIRNOW PM10 and Pm25 data

Link to dust modelers for faster collective
learning?
41
Monte Carlo simulation of dust transport using
surface winds (just a toy, 3D winds are
essential!)

• See animation Note, how sensitive the transport
  direction is to the source location (according to
  this toy)