| HOME | KCARTA | LBL | PLASMA | GENERAL |
Line-by-line codes to compute optical depths usually take a long time to run, but have to be accurate as many models are developed using these codes. I have written a Matlab/Fortran code to compute spectra for the various molecular spieces in the Earth's atmosphere, using spectral line parameters from any of the HITRAN (1992, 1996, 1998, 2000) databases. Most of the code is fairly standard, and is based on Dave Edwards' GENLN2 code to compute the spectra. However, the UMBC-LBL code is a little more careful when including the effects of lines on the region of interest : lines are binned into three : near, far and medium distance away from region of interest.
This code can be used to generate the "monochromatic" spectral lineshapes at various atmospheric states (temperatures, pressures and gas amounts). Using this, new kCARTA compressed databases can be produced when needed (such as when a new HITRAN database is released). While the LBL code takes a long time to run, and so generate a kCARTA database, this is irrelevant as far as users of kCARTA are concerned, as they just need the compressed database that we give out. As mentioned above, kCARTA can then be very quickly used to generate transmittances for model atmospheres, from which users can develop fast forward models for various instruments.
I have also developed two special versions of this code, for computing the spectra of CO2 and H2O in the infrared region. Most of the base work here has been done by a former UMBC PhD student Dave Tobin (now working at UW-Madison).
For CO2, the line by line code will use our latest estimates for CO2 line mixing parameters, as well as duration of collision effects, in both the 4 um and 15 um regions. Because the transmittances in these regions vary from 0 to 1 over a small spectral range, they are used as sounding channels to give information about the atmosphere at various altitudes. An accurate lineshape model in this region is therefore crucial.
An example of the importance of doing CO2 line mixing is shown in the next figure, at 2390 cm-1. The plot shows error estimates from NAST data, as well as comparisons to GENLN2, line mixing with birnbaum, and line mixing with birnbaum and cousin chi functions. The best fit to the data is the last computation, shown in red. The plot in magenta is the GENLN2/Cousin lineshape, which is the model used currently. Our model reduces the errors in brightness temperature by upto 2K.
New laboratory data we have just recieved, as well as comparisons of our model with older models for various other instrument campaigns (CAMEX, WINTEX etc), do indicate that we have indeed made a big improvement in the 4 um region, and that there are small but significant improvements in the 15 um region.
Examples of comparisons with one of the NAST sounding campaigns can be found here .
We have also obtained new water vapor data from Rutherford Appleton Lab in
England, which we will begin analysing when time permits. In the meantime, the
line by line code written for H2O can use the local lineshape multiplied by chi
functions, and add in one of four CKD continuum versions (CKDv 0, 2.1, 2.3 or
2.4). Looking at campaigns such as CAMEX and WINTEX, suggests that the
CKD 2.4 continuum in the 1585-1615 cm-1 bandhead is too low; a more
reasonable estimate for the foreign continuum in that region is the CKD
0.0. The RAL continuum data we have suggest that the foreign coefficients
are bounded by CKD 0.0 and CKD 2.4; consistent with the campaigns, CKD 0
seems a better fit near the 1600 cm-1 bandhead, while CKD 2.4 gives
better results away from the bandhead. However, at 1600 cm-1 the
CKD 0, 2.1, 2.3 and 2.4 self coefficients also seem to be incorrect
Back to top