NWP SAF: RTTOV v6 bugs

Last updated 11/4/2001

1. The infrared Line-by-Line transmittances from which the RT coefficients are computed did not asymptote to 1 at the top level but to 0.998 due to an error in the convolution of the HIRS and other IR channel transmittances. This results in small forward modelling errors (<0.1K for HIRS channels) and small unphysical increases in the temperature Jacobians dBT/dT(p) at the top layer (0.1hPa to space). This does not affect the microwave channels. The code is not in error only the coefficient files for any infrared channels. It is planned to replace the coefficient file for ATOVS with correct coefficient files when the file is updated to include NOAA-16 coeffs.

2. Erik Andersson at ECMWF has pointed out there is a bug in the emissad.F and emissk.F routines so this only affects those of you using the adjoint or K versions of the code not the forward model. In emissad.F and emissk.F the last line of that routine should be:

(not EMIS(J) = EMIS(J) + PEMS(J) )

The problem is that the array EMIS has not been zeroed before the accumulation of gradient contributions starts. It gave incorrect gradients of Jo w.r.t. emissivity. You are advised to change your version of the code to remove this error.

3. There is an error in the coefficients for the infrared surface emissivity model for AVHRR channels only. AVHRR channel 5 (12 micron) is the most affected with emissivity differences of up to 0.01 (giving brightness temperature differences of ~0.1K). The other AVHRR channels are only affected at the 0.01K level. Note if you specify the surface emissivity for the channels there is no error introduced. A corrected coefficient file (which now includes NOAA-16 coefficients) is available - please send a request using the NWP SAF Feedback Form.

4. In subroutine rttovk.f variable CLDW is not initialised so increments cloud absorption on each call of rttovk.
Add after line 176: CLDW(:,1:KNCHPF)=0.

5. If you want to exit with an error code when the opening of the coefficient file fails, you have to replace the stop statement inside abor1.F by "call exit(1)". This should only inform users on how the code manages the open failure.

6. For SSM/I and TMI the fractional cloud cover in the PCV array should always be set to zero. There is a bug which treats the affect of cloud on SSMI/TMI radiances like the IR radiances if fractional cloud cover is not zero. For AMSU radiances the code forces the cloud cover to be always zero. The effect of cloud liquid water on the SSMI/TMI radiances is provided for by filling the cloud liquid water profile array PAV(;,4,:) with non zero values.