Reading IASI L1C data (EC)
[Prev]
[Next]
Comparing BEAT software with IDL routine obtained from EUMETSAT.
- BEAT is approx 5 x faster but requires local installation.
- Brightness
temperature spectra differ, but only in last significant figure.
- Cosine of the zenith
angle differ by up to 2%.
- BEAT only has time resolution of 1sec, but EUMETSAT
routine in msec.
Reading IASI L1C data (AS)
Bug found in ECMWF code for reading IASI data from BUFR files,
seems to be associated with Gfortran compiler (reported to ECMWF).
IASI NRT web-site (CH)
[Prev]
[Next]
Code now transferred to CEDA and trying to get it to run there.
Will try to define minimum netCDF data structure for externally-supplied
data to be added to web-site (eg value, lat, lon, time, pixel#, scan-position,
flags, ...). Expects one file per orbit.
Retrieval Levels (TV)
[Prev]
[Next]
Continues to work on paper on optimal selection of retrieval levels for
IASI. CH4 averaging kernels show surprising sensitivity of tropospheric CH4
to high altitude CH4.
Minor fixes to IASI noise covariance matrix calculation procedure.
RAL Retrievals (RS)
[Prev]
[Next]
Finishing off a couple of EUMETSAT studies.
Diane Knappett's start postponed until January. She will be working
on total column CH4 retrievals from the IASI 3.7um band.
RFM (AD)
[Next]
Working on speeding up RFM simulations of IASI data.
Two approaches:
- Use pre-tabulated Look-Up Tables of absorption coefficient k(v,p,T)
- Use irregular grid for radiative transfer
For tests, infrared spectrum split into 500cm-1 bands from 5000-3000cm-1,
US standard atmosphere (50 layers), surface temperature +10K relative to
lowest atmospheric temperature, 6 absorbers: CO2, H2O, O3, CH4, N2O and CO.
Line-by-line spectra calculated at 0.001cm-1 spacing then convolved with
IASI AILS to 0.25cm-1 spacing. RFM compiled with ifort using the '-p' option
(and gprof program) to assess CPU time.
Plots show (top) LBL spectrum, (bottom) difference wrt LBL spectrum
of
- LBL_NOCHI: line-by-line but without CO2 Chi-factor and line-mixing
- LUTV: Using look-up tables interpolated to regular 0.001 cm-1 grid
- GRD: LBL performed on irregular grid (implicitly interpolated
during AILS convolution)
- LUTGRD: Using look-up tables on their 'natural' irregular grid determined
by spectral structure within each LUT
- TABGRD: Using LUTs generated only at same irregular grid points used for
the radiative transfer
Figures in legend give SD of differences.
| CPU time (sec) to generate RFM spectra
|
|---|
| Band | LBL | LBL_NOCHI | LUTV | GRD | LUTGRD | TABGRD
|
|---|
| 500-1000 | 776 | 293 | 46 | 110 | 93 | 5
| [Plot]
|
| 1000-1500 | 359 | 282 | 45 | 50 | 93 | 5
| [Plot]
|
| 1500-2000 | 251 | 187 | 42 | 17 | 87 | 2
| [Plot]
|
| 2000-2500 | 780 | 388 | 50 | 93 | 96 | 6
| [Plot]
|
| 2500-3000 | 297 |
253 | 47 | 19 | 96 | 2
| [Plot]
|
