investigate potential spurious jacobians

[BenjaminTJohnson]

from Eric Maddy (NOAA-STAR):

Jacobians are non-zero (and "noisy") where there should be little or no response.

See attached figures which show CRTM Jacobians(channel, pressure) for MetOp-C IASI B1, B2, B3 and US STANDARD TROPICAL profile. Temperature is d(brightness temperature)/dT while other gases are d(brightness temperature)/d(log(VMR))

A few weird things:
Note the "blips" in CO2, CH4 in B2 for frequencies > 1400cm-1.
CH4 responses at 2400cm-1.

I seem to recall that the spurious non-zero Jacobians were larger (2x-3x), but that was forever ago, and I was looking at a different CRTM and different input profiles. I've also compared these to CrIS FSR simulations. Those look similar with generally smaller magnitudes due to wider channel responses.

I've also attached figures of vertical Jacobians for all channels plotted on top of each other. Many of the CO2 and CO Jacobians in B2 look very noisy. I suspect this is related to the issue of using non- or extremely weak absorbing gases to predict transmittances in those channels.

[quanhualiu]

Eric, there are more than 200,000 absorption lines between 2000 cm-1 and 2600 cm-1 for H2O, CO2, O3, N2O, CO, CO, and CH4. It seems that the CRTM Jacobian features match with absorption lines. CRTM team needs to look into line by line Jacobian calculations to determine whether some noise results due to the uncertainties in the CRTM calculations.