Evan A' Ellicott1, Eric F' Vermote1, Francois Petitcolin2, and Simon J' Hook3

1
Validation of a New Parametric Model for
Atmospheric Correction of Thermal Infrared
Channels
Evan A. Ellicott1, Eric F. Vermote1, Francois
Petitcolin2, and Simon J. Hook3 1Department of
Geography, University of Maryland, USA 2ACRI-ST,
France 3JPL, USA 1Contact info eaelupus_at_umd.edu
AGU 1551
SST Evaluation
Sensitivity to Atmospheric Profile
The eventual goal is to employ the parametric
model as part of an operational atmospheric
correction scheme for MODIS. With this in mind
we included an analysis to test several sources
of atmospheric profile data. We developed a
reference data set using radiosonde profiles
and MODTRAN. The parametric model was then tested
against the reference using MODIS and AIRS
profiles (Figure 8).
The parametric model accuracy was tested against
the Aqua-MODIS sea-surface temperature product
(MYD28-SST) (Figure 3). The split-window
approach generally used for retrieving SST is
accepted to be fairly accurate over water targets
(Eugenio et al., 2005), due in part to stable
target emissivity, thus offering a bench-mark to
compare against. Results from multiple
observations show a good fit (Figure 5).
Overview Goals

• Surface temperature is a key component for
  understanding energy fluxes between the Earths
  surface and atmosphere, and therefore a critical
  part of climate prediction, analyzing vegetative
  stress, and hydrologic modeling. Accurate
  retrieval of surface temperature from satellite
  observations requires proper correction of the
  thermal channels for atmospheric emission and
  attenuation. Although the split-window approach
  has offered a relatively accurate method, this
  empirical approach is subject to bias (Minnett,
  2003). Single channel correction reduces
  uncertainty inherent to the split window method
  and preserves spectral signatures in thermal
  infrared, but requires an accurate radiative
  transfer model and detailed description of the
  atmospheric profile. Computation time is
  generally a limitation of radiative transfer
  models for operational use.
• We present here a thermal parametric model based
  upon the MODTRAN radiative transfer code and
  tuned to MODIS channels. The goal was to assess
  the accuracy of the parametric model against
  MODTRAN and then with various sources of surface
  temperature data. In addition, we investigated
  several sources of atmospheric profile data a
  critical input variable to radiative transfer
  modeling for accurate atmospheric correction.

Fig 3. MODIS SST product (MYD28) for the western
United States and Pacific Ocean 12 August 2003.
Cloud cover is distinguished by the dark
blue-black.
Fig. 4. Comparison of the parametric model and
Aqua-MODIS (MYD28) SST. Near-nadir observations
are from 2004 (N 82 ) blue points correspond
with DOY337 at 1725 UTC magenta with DOY336
at 0150 UTC and green with DOY012 at 0845 UTC.
a.
b.
Fig.8. MODIS band 31 temperature comparison
between the radiosonde-MODTRAN derived
reference and parametric model derivation using
(a) AIRS or (b) MOD07 profiles.
in situ Validation
Ground based temperature measurements using in
situ radiometers offers the opportunity to fully
validate the accuracy of our parametric model
against real temperatures that have little to no
influence from atmospheric perturbation.
Comparisons were made between both water and land
surface targets. Water surface observations were
made over Lake Tahoe, Nevada, using radiometers
located on 4 permanently moored buoys (see Hook
et al., 2003) (Figure 5). Validation with land
surface temperature observations was made with
data provided by Coll et al. (2005).
Measurements were made with tripod mounted
radiometers placed over stable, homogeneous sites
in eastern Spain (Figure6).
Synthetic Dataset Analysis
1. Performance of the parametric model
correction for atmospheric upward and downward
radiances and atmospheric transmittance was
assessed against MODTRAN using NCEP atmospheric
profiles and sun/sensor geometric data from Terra
MODIS granule 2001.047.0800.
Fig. 9. The larger error in the MODIS profile
seen in Figure 8 can be attributed to lower
spectral resolution as compared with AIRS.
Higher columnar water vapor content recorded by
MODIS led to an overcorrection and thus bias
towards overestimation of temperatures. This
graph shows an example of the variation between
profile sources.
Fig. 5. Bathymetric (100m) map showing locations
of in situ skin and bulk temperature observations
of Lake Tahoe, NV (TR1 TR4). A raft at each
location contains a radiometer mounted 1m above
the surface making continuous observations in the
7.8-13.6µm range. Additional temperature sensors
at several depths record bulk temperature.
Conclusion
c.
a.
b.
The parametric model offers a method to
operationally correct the at-sensor radiance
values for atmospheric perturbations. Although
computational time was not measured as a part of
this analysis, anecdotal comparisons between the
parametric model and MODTRAN suggests at least an
order of magnitude increase in speed.
Consideration of the profile used in any
radiative transfer model is paramount to
achieving accurate correction for atmospheric
effects. This study highlights a variety of
potential profile sources and demonstrates a
relatively good accuracy, especially when
compared with in situ temperature data.
2. With TOA radiance, surface emissivity, and the
atmospheric parameters retrieved above we
calculated surface radiance 1. The Plank
function is inverted to calculate surface
brightness temperature 2. 1
2 Bending of the optical path when
view angles exceed 60 is accounted for in
MODTRAN , but not in the parametric model
therefore we excluded observations above 60.
Fig. 1. Comparison of MODTRAN and the parametric
model derived (a) atmospheric upward radiance,
(b) atmospheric downward radiance, and (c)
atmospheric transmittance using synthetic
dataset. N423.
Fig. 6. Study sites used for measurement of LST.
In situ data were recorded along transects within
flat, homogeneous plots consisting of cultivated
rice fields. Courtesy of César Coll.
Literature Cited
Fig. 7. A good agreement between surface
temperature derived from the parametric model and
in situ measurements. The blue diamonds
correspond with Lake Tahoe measurements. Pink
squares correspond with rice field LST
measurements.
Coll, C., Caselles, V., Galve, J. M., Valor, E.,
Niclos, R.., Sanchez, J. M., Rivas, R.
(2005). Ground measurements for the validation of
land surface temperatures derived from AATSR and
MODIS data. Remote Sensing of Environment, 97,
288-300. Eugenio, F., Marcello, J.,
Hernandez-Guerra, A., Rovaris, E. (2005).
Regional optimization of an atmospheric
correction algorithm for the retrieval of sea
surface temperature from the Canary
Islands-Azores-Gibraltar area using NOAA/AVHRR
data. International Journal of Remote Sensing,
26, 1799-1814. Hook, S. J., Prata, F. J., Alley,
R. E., Abtahi, A., Richards, R. C., Schladow, S.
G., Palmarsson, S. O. (2003). Retrieval of lake
bulk and skin temperatures using Along-Track
Scanning Radiometer (ATSR-2) data A case study
using Lake Tahoe, California. Journal of
Atmospheric and Oceanic Technology, 20,
534-548. Minnett, P. J. (2003). Radiometric
measurements of the sea-surface skin temperature
the competing roles of the diurnal thermocline
and the cool skin. International Journal of
Remote Sensing, 24, 5033-5047.
a.
b.
Fig. 2. Surface brightness temperatures
comparison between the parametric model and
MODTRAN. Emissivity was set to (a) 1.0 and (b)
0.9. N423