| Gastropod (Sherlock, 2001b) is a fast
radiative transfer code developed to meet the requirements for a
day-one radiative transfer operator for AIRS (and eventually IASI)
for use in variational data assimilation systems.
This code development was undertaken for three main reasons
- While the existing AIRS fast model, PFAAST
(Hannon, 1996), has good forward model error characteristics,
the corresponding adjoint (k) model - required in variational
data assimilation schemes -has not been developed, nor is there
any plan to do so. Thus PFAAST cannot be integrated into a variational
assimilation/retrieval package as it stands.
- An alternative solution, modification of the existing IASI
fast model RTIASI (Matricardi, 1999), was not adopted because
of concerns relating to the robustness and the complexity of
this code in its current form (regression instability, code
complexity due to explicit treatment of negative transmittances).
- Any new fast model development benefits from past experience.
Thus the Gastropod development exploits advantageous methods
identified in a previous fast model intercomparison study (Sherlock,
2000) and a subsequent study of the vertical discretisation
requirements for advanced sounder fast radiative transfer algorithms
(Sherlock, 2001a). Model simplicity - facilitating both the
coding of gradient routines, and the understanding of model
error characteristics - has been favoured wherever possible.
The Gastropod forward model formulation is based principally
on the PFAAST AIRS model developed by Hannon (1996) (note model
methodology has been extended to IASI). In particular, separate
prediction of water vapour line and continuum absorption has been
adopted based on the performance of the PFAAST model in the H2O
band and the longwave window regions. However, the calculation
of layer mean parameters for the effective optical depth prediction
scheme and the radiative transfer (RT) calculation has been modified.
The new estimation of layer mean quantities uses a formulation
which has a well-defined adjoint mapping from RT layer mean quantities
to input profile variables in all situations. Similarly, a single
case prediction scheme for water vapour line absorption has been
adopted so as to avoid the errors (discontinuities) in modelled
Jacobians associated with the transition between regression schemes
identified in Sherlock (2000). Finally, GASTROPOD makes an important
extension to the PFAAST model, as tangent linear and adjoint (k)
models have also been coded and validated.
Hannon,S., L. Strow, and W. McMillan, 1996. Atmospheric infrared
fast transmittance models: a comparison of two approaches. In
Optical Spectroscopic Techniques and Instrumentation for Atmospheric
and Space Research, pp. 94-105.
M. Matricardi and Saunders, R.W., 1999. A fast radiative transfer
model for simulation of IASI radiances. Applied Optics
Sherlock, V.J., 2000: Results from the first UKMO IASI radiative
transfer model intercomparison, Numerical Weather Prediction
Division Technical Report No. 287.
Sherlock, V.J., 2001a: Vertical discretisation for advanced sounder
radiative transfer models: grid refinement for RTIASI. Numerical
Weather Prediction Division Technical Report No. 336.
Sherlock, V.J., 2001b: Gastropod: A fast radiative transfer model
for AIRS and IASI, To appear as a Numerical Weather Prediction
Division Technical Report.