Miguel A' Bustamante

1
Improved Processing of Multi-filter Rotating
Shadowband Radiometer Network for Distributed
Monitoring of Atmospheric Aerosols

• Miguel A. Bustamante
• Mentors and Collaborators
• B. Gross, F. Moshary, and S. Ahmed The Graduate
  Center, Optical Remote Sensing Lab, NOAA-CREST,
  The City College of the University of New York
  and NASA-GISS

With the kind assistance of Dr. M. Alexandrov
(NASA-GISS)
2
Can Satellites be used for examining air quality
in urban areas

• Aerosol optical depth from satellites show strong
  correlation with PM2.5 measurements on ground in
  the North East
• Aloft Plumes (as seen from Lidar) must be
  filtered out.
• Satellite results cannot provide the spatial
  resolution needed to accurately quantify the
  local aerosol signal generated within the urban
  canopy
• Development of a spatial network of radiometers
  can help improve modeling of urban albedo and
  provide a subsequent improvement in spatial
  resolution.

3
Aerosol optical depth from satellites shows
strong correlation with PM2.5
Coincident Lidar measurements have been used to
filter out significant aloft plumes from dataset.
IDEA 60ug line
AOD for High PM 2.5 region can easily give
false positives
PM2.5 Concentration
Still, we note poor performance of AOD as a
predictor of high pollution events due to errors
in retrieval and aerosol inhomogeneity
4
Satellite Observations inferring larger aerosol
loading over city
Note that measurements over NYC have been
derived based on regional algorithm which
removes bias
5
MFRSR Ground Radiometer Network Motivation

• Satellite AOD does correlate well with PM2.5 but
  in general can lead to many false positives.
  Spatial resolution is in general too low to
  accurately mark the boundary.
• Such measurements can only reliably provide
  statistical information on aerosol differences
  between urban and non urban regimes
• Need a cost effective spatially distributed
  radiometer system which can validate satellite
  measurements over broad spatial distributions.
• Unfortunately CIMEL radiometer are too expensive
  and not portable vs. a cheaper and portable MFRSR
  instrument.
• The network requires a robust processing approach
  which is less sensitive to instability in AOD
  than the Langley Regression Calibration.
• Besides monitoring aerosol transport and
  validating satellite measurements, these
  measurements can be used to improve the retrieval
  algorithms over different areas by improving
  surface albedo.

6
Devices
CIMEL Sun Photometer CE-318

• Multi-filter Rotating Shadowband
• Radiometer (MFRSR)

It makes direct sun measurements at eight
spectral bands 340, 380, 440, 670, 870, 936 and
1020 nm. The CCNY sun photometer is identified at
NASA AERONET (AErosol RObotic NETwork) as
instrument 237. Since light is absorbed and
scattered by atmospheric anthropogenic gases, the
concentration of Aerosols and NO2 among others
can be determined.
Model MFR-7 is a field instrument that
simultaneously makes instantaneous spectral
measurements at six wavelengths (415, 500, 615,
673, 870, and 940 nm) of Global, diffuse and
direct normal components of spectral solar
irradiance. The instrument uses the same detector
to sense global and diffuse irradiance
eliminating channel variability.
7
Network TopologyCurrent and Future
Outer ring
Inner ring
8
Data Analysis(1)

• Improvement on this approach are possible by
  implementing a novel algorithm based on the ratio
  between the direct and diffuse radiance developed
  at NASA-GISS in which only the optical depth
  ratios during the calibration procedure are
  required to be stable.
• Results shows that this approach significantly
  improve optical depth time series measurements
  when compared to Aeronet CIMEL aerosol optical
  depth measurements in comparison to the Langley
  regression method calibration.

Regressing these equations to find D and C870. D
is the Opacity Deficiency Term

• M. Alexandrov, A. Lacis, B. Carlson and B.
  Cairns, 2002 Remote Sensing of Atmospheric
  Aerosol and Trace Gases by Means of Multi-filter
  Rotating Shadowband Radiometer. Part I J. Atmos.
  Sci., 59, 524-542

9
Data Analysis(2)

• Once the calibration of the 870 channel is
  accomplished, the calibration of the other
  aerosol channels can be accomplished with 870
  optical thickness as reference.
• Using the 870 as reference leads to a regression
  problem where it is only necessary to assume that
  the aerosol optical depth ratio between the
  channel and the 870 reference is stable.
• This condition is much less severe than the
  condition of the Langley regression which
  requires the aerosol optical depth itself should
  be stable.

10
Comparison Between Uncertainties in AOD and AOD
Ratio
Note that the fractional error of the aerosol
optical depth is at least twice as large as the
Aerosol Optical depth Ratio which supports the
effort to use optical depth ratios in the
calibrations stage.
11
Comparison Between CIMEL and MFRSR
NASA processing algorithm matches CIMEL process
more robustly
12
Further Match-ups to Assess When MFRSRBreaks
Down
Problems at low AOD
Showing worst case errors for NASA Langley and
other technique
13
Further Match-ups to Assess When MFRSRBreaks
Down
Showing worst case errors for NASA Langley and
other technique
14
Preliminary Results fromPrinceton Site
Effect of a Dust Plume on Coarse versus Fine Mode
Aerosol
MFRSR Coarse and Fine Mode Optical depth sees
dramatic increase in Coarse particles at 1600
Three Channel Lidar sees dust incursion into PBL
at 1600 and later
15
Match-ups of AOD (including fine and coarse mode)
at DOE ARM Site in Southern Great Plains (SGP)
In general, retrieval of aerosol fine mode is
superior than coarse mode due to lack of long
wavelength channels (wavelength lt 860)
16
Conclusions

• In general processing MFRSR data using Langley
  regression leads to significant errors due to the
  variability of the optical depth during the
  calibration process and difficulties in the
  elimination of sky background.
• An improved processing method based on
  calibration of the long channel (870nm) using a
  combination of direct and diffuse radiation does
  not require good stability. Calibration of the
  other channels assumes that the optical depth
  ratios are stable which we showed is a realistic
  assumption.
• We note that good agreement often occurs (not
  always) if CIMEL AOD is high and variability
  with the angstrom coefficient is low
• A comparison of CIMEL and MFRSR in the retrieval
  of fine and coarse mode aerosols indicates that
  the fine mode retrieval by MFRSR seems much
  better than the coarse mode. This can be
  partially explained by remembering the MFRSR
  wavelength are in the range from 440 to 870nm
  unlike the CIMEL which goes from 340 to1020nm (or
  more recently 1640nm)

17
Future Work

• More instruments are being networked into the
  system.
• Intercomparisons to explore case by case
  differences in AOD between different sites are
  needed and will be based on situations where
  clear down wind or cross wind conditions between
  sites
• More data is needed to quantify the retrieval
  performance for total and fine mode AOD
• Data from GISS algorithm needs to be implemented
  over network based on CART Architecture.

18
Thank You