Retrieval of smoke aerosol loading from remote sensing data

1
Retrieval of smoke aerosol loading from remote
sensing data

• Sean Raffuse and Rudolf Husar
• Center for Air Pollution Impact and Trends
  Analysis
• Washington University

2
Overview

• Problem statement and goal
• Method
• Radiative transfer theory
• Aerosol map generation
• Summary
• Continuing work

3
Problem statement and goal
Problem

• Biomass burning contributes a significant
  fraction of the anthropogenic aerosol
• Wildfires and prescribed burns
• Slash-and-burn agriculture
• Crop waste burning
• The amount of aerosol generated by biomass
  burning is not well quantified
• No satisfactory tracer for biomass smoke has been
  found
• Ground and aircraft-based studies do not provide
  adequate spatial coverage
• Aerosols from smoke contribute to global cooling
• Quantification is needed to model global climate
  change

Goal
To quantify the emission of smoke from biomass
burning as well as study its spatial and temporal
pattern
4
Method remote sensing of aerosol optical
properties

• Remote sensors deployed in research satellites
  detect radiation from the earth and its
  atmosphere
• These sensors allow us to detect aerosols that
  scatter and absorb light
• We utilize the Sea-viewing Wide Field-of-view
  Sensor (SeaWiFS) instrument on NASAs SeaStar
  spacecraft
• Polar-orbiting
• 1 km resolution
• Daily coverage
• 8 channels (6 visible, 2 near-IR)

5
Radiative transfer theory for aerosol-surface
co-retrieval
The sensed radiation is decomposed into
scattering and absorption by (1) gases, (2)
aerosols as well as reflection from the (3)
surfaces and (4) clouds. Air scattering and
surface/aerosol reflectance are assumed to be
additive, disregarding multiple scattering
effects.
6
Apparent surface reflectance, R

• The surface reflectance R0 objects viewed from
  space is modified by aerosol scattering and
  absorption.
• The apparent reflectance, R, is R (R0 Ra)
  Ta

Aer. Transmittance Both R0 and Ra are attenuated
by aerosol extinction Ta which act as a filter
Aerosol Reflectance Aerosol scattering acts as
reflectance, Ra adding airlight to the surface
reflectance
Surface Reflectance The surface reflectance R0
is an inherent characteristic of the surface
Apparent Reflectance R may be smaller or larger
then R0, depending on aerosol reflectance and
filtering.
Aerosol as Filter Ta e-t
Aerosol as Reflector Ra (e-t 1) P
R (R0 (e-t 1) P) e-t
7
Obtaining aerosol optical thickness from excess
reflectance
The perturbed surface reflectance, R, can be used
to derive the the aerosol optical thickness, t ,
provided that the true surface reflectance R0 and
the aerosol reflectance function, P are known.
The excess reflectance due to aerosol is R- R0
(P- R0)(1-e- t) and the optical depth is
As R0 increases, the same excess reflectance
corresponds to increasing values of t. Accurate
and automatic retrieval of the relevant aerosol P
is a difficult part of the co-retrieval process.
Iteratively calculating P from the estimated t(
?) is one possibility. t can be related to mass
loading by assuming physical and optical
properties.
8
Aerosol effects on surface colorandSurface
effects on aerosol color
The image was synthesized from the blue (0.412
µm), green (0.555 µm), and red (0.67 µm) channels
of the 8 channel SeaWiFS sensor. Air scattering
has been removed to highlight the haze and
surface reflectance.
9
Process for co-retrieval

1. Generate daily total reflectance image with air
   reflectance removed, R
2. Generate surface reflectance image, R0
3. Subtract daily total reflectance image from
   surface reflectance image to get aerosol optical
   thickness, t
4. Filter t, removing clouds and other interferences

10
1. Daily reflectance image

• 2000-08-23 RGB image after preprocessing
• Preprocessing includes
• Conversion from L1a engineering values to L1b
  scientific values (counts ? radiance)
• Georeferencing
• Splicing
• Rayleigh correction

11
2. Generating the surface reflectance, part 1

• The surface image is the clean surface image
  with all clouds, air, and aerosol removed
• Daily surface reflectances are generated by
  creating a composite image from the nearest 15
  days
• At each pixel, the cleanest daily value is used
• As aerosol and clouds both make the reflectance
  brighter, the cleanest value is the one with the
  lowest reflectance
• Cloud shadows and other anomalous low values are
  not used

12
2. Generating the surface reflectance, part 2
Uncleaned Surface Reflectance

• In 15 days, some locations are not cloud and
  aerosol free
• This results in leftover haze, and areas of
  continual cloud cover
• We use a small (15-day) time span to preserve
  temporal surface change, such as in the fall
• However, the blue channel remains fairly constant
  over a longer time period
• Leftover aerosol signal is subtracted from a
  60-day blue minimum
• Other channels are subtracted assuming a
  wavelength dependence of t

Cleaned Surface Reflectance
13
3-4. Generating aerosol optical thickness (t)

• Aerosol optical depth (t) is then calculated from
  the daily total reflectance (R) and surface
  reflectance (R0)
• Clouds are removed using several filters based on
  the spectral characteristics of t
• This image shows the blue channel (412 nm)
  aerosol optical depth

14
Total reflectance and optical depth comparison
Haze
Smoke plume
Filtered clouds
15
Summary

• Biomass smoke is difficult to quantify
• No satisfactory tracers have been discovered
• Ground-based and aircraft studies do not provide
  good spatial coverage
• Aerosol optical thickness can be retrieved from
  remote sensing imagery
• With knowledge of particle physical and optical
  properties, an estimation of mass loading can be
  made
• Size distribution, morphology, mixing regime
• Extincion coefficient, single-scatter albedo,
  phase function

16
Continuing work

• Estimation of smoke fluxes
• Identify specific smoke plumes
• Divide map into location grids
• Use wind vector data to calculate flux through
  the grids
• These values are required for climatological
  models
• Data fusion
• Data from remote sensing and ground-based
  networks are complimentary
• Multiple data sets will be fused to improve
  understanding

17
Thank You!

• R. Husar, F. Li, E. Vermote
• M. King, Y. Kaufman, D. Tanre, J. Martins, P.
  Hobbs . . .
• U.S. EPA