The satellite sensor measures the infrared radiance leaving the top of the
atmosphere towards the satellite; this radiance is corrected with respect
to the influence of a clear (i.e. nearly non scattering) atmosphere; the
resulting radiance is converted to a temperature according to Planck's
law and it is called "Land Surface Temperature" (LST).
The LST is driven by the incoming solar radiance and by the long-wave
irradation from the atmosphere. It is related to the outgoing terrestrial
infrared radiation, the sensible and latent heat flux, and the surface
heat flux. As the LST is an essential parameter of the energy balance
at Earth's surface,
ONC needs
heating rate (i.e. gradient of LST from sunrise to noon) on a hourly basis.
OLF will test
the impact of a coarse resolution LST product on the reliability of its
models of change detection, since 1 km resolution LST is not available.
OFM uses the LST
from EWBMS database in crop monitoring and yield forecasting studies.
The University of
Karlsrhue (IMK) and
EARS assess the
LST from METEOSAT images using different approaches.
LST from METEOSAT-7 images, August 2000.
IMK derives LST from METEOSAT thermal infrared channel images using a
neural network. METEOSAT is the only satellite that provides infrared
measurements over Africa and Europe that resolves the diurnal wave of
LST. The disadvantage of METEOSAT is that only one infrared
channel is available. Thus established LST determination methods like
split-window technique can not be used and the atmospheric state can
not be derived from METEOSAT data. The atmospheric situation
(i.e. the temperature and moisture profiles) are taken from ECMWF analysis
or re-analysis. The physics of the atmospheric correction of a single
infrared channel is the following: calculate the
expected satellite measurement for a reasonable range of land surface
temperature, surface elevation and emissivity for the actual profiles
and viewing angle - this constitutes the forward calculation of
atmospheric radiances. The LST is then determined by interpolation to
the actual satellite measurement (this is the inversion of the forward
calculation) for profiles around the current pixel and horizontal
interpolation of the atmospheric correction at the surrounding pixels.
A neural network is used to speed-up this procedure. Its imputs are: the
actual satellite measurement, the temperature and moisture profiles, the
surface elevation and emissivity, as well as the viewing angle. The LST
is not estimated for cloudy pixels. IMK has developed a cloud detection
for the three spectral channels of METEOSAT based upon temporal
development (bright surface is separated from clouds due to the duration)
and sophisticated threshold techniques (IR thresholds for each slot are
determined dynamicaly from several days up to a month, depending on the
cloudiness). The resulting cloud mask is also used by
CNRM/Météo-France
for
DSR retrieval and by
IM for
DLR assessment.
10-day LST from METEOSAT images over the
Euro-Mediterranean region for August 2004 (left),
and over the West Africa region for December 2002.
EARS provides the Land Surface Temperature generated by the
EWBMS
(Energy Water Balance Monitoring System) from METEOSAT noon and midnight
thermal infrared (10 - 13.1 µm) images,
which yield, after calibration, the planetary noon and midnight
temperature. These latter are converted to land surface temperatures by
means of an atmospheric correction procedure. The atmospheric correction
coefficient is determined using a reference within the image. The
reference is obtained from the driest pixels, which have the highest
planetary temperature over global radiation ratio. It is assumed that
these pixels have zero evapotranspiration and thus sensible heat flux
equals net radiation. For these pixels, the land surface temperature may
then be calculated. The resulting pair of highest planetary and land
surface temperature allows to calculate a first order atmospheric
correction coefficient, which is then applied to all other pixels of the
noon and midnight images. The average surface temperaure is obtained by
averaging the noon and midnight surface temperature.
References
