15.
Author：
An innovative physical scheme to retrieve simultaneously surface
temperature and emissivity from IASI
Maxime Paul, F. Aires, C. Prigent
Affiliation： LERMA
The retrieval of key atmospheric parameters such as water vapor or temperature
close to the surface from satellite measurements is difficult to perform over land
areas. A precise knowledge of the state of the surface is needed in order to interpret
as precisely as possible its interaction with the measured radiation. A way to
characterize this interaction is to determine the surface emissivity and skin
temperature. This study focuses on the infrared wavelength domain. The objective is
to create a retrieval algorithm based on infrared measurements (here IASI instrument
will be used) that gives real time access to the surface emissivity and skin
temperature. This algorithm uses as inputs an emissivity and surface temperature
first guesses and IASI measurements. It retrieves directly both the effective
emissivity and the surface skin temperature.
To get the emissivity first guess required by the retrieval algorithm, a spectral
emissivity interpolator has been built. It is based on MODIS emissivity retrievals and
laboratory measurements at IASI spectral definition. A non-linear method (neural
network) has been used to better represent the fine spectral features. To reduce the
time of the calculations, a principal component analysis is performed over the high
spectral emissivity spectra. This dataset has been developed to provide an emissivity
first guess compatible with the principal component analysis used in the retrieval
algorithm.
A retrieval scheme is built using IASI measurements, the interpolated emissivities
and corresponding ECMWF analysis as inputs and the surface temperature and the
effective emissivity spectrum as outputs. Its principle is based on a local
linearization of the radiative transfer equation. The interpolated emissivities and the
ECMWF analysis are used to perform a radiative transfer calculation to get a first
guessed IASI spectrum used by the retrieval.
In order to measure the precision of this algorithm an a posteriori method is adopted.
It consists in comparing radiative transfer simulated data using different emissivity
datasets as inputs, combined with ECMWF analyses. The infrared emissivity
datasets come from different sources: NASA (D. Zhou), UWIREMIS (E. Borbas)
and the ARA group. Such comparisons show the importance of a precise emissivity
knowledge. The use of a monthly mean emissivity climatology instead of a real time
retrieval reduces the quality of the surface characterization (by up to 5K in the 1000
to 1100 cm-1 band) and thus can induce errors in the following atmospheric
retrievals. Indeed, as this surface characteristic retrieval will be used as a first step
before atmospheric parameter inversions, if its results are not precise enough,
mischaracterization of the surface will imply errors in the atmospheric state.