Chris successfully defended his PhD dissertation on November 29, 2011.
Multivariate Retrieval of Carbon Monoxide
A new technique is presented here to retrieve carbon monoxide (CO) profiles from Atmospheric Emitted Radiance Interferometer (AERI) spectra. This retrieval version deviates from the previous AERI CO retrieval method, which utilized signal processing to determine a constant CO mixing ratio representative of the entire troposphere. Instead, this retrieval version utilizes linear mapping to ascertain an estimate of the CO profile. A detailed analysis is conducted to estimate the error from all aspects of the the linear mapping procedure including measurements, forward modeling of atmospheric radiation, and uncertainty from inputs to the forward model. It was found that the dominant sources of error were from cloud contaminated spectra and uncertainty in absorption line strengths inside the forward model, A new cloud flagging technique that uses a neural network to identify spectra affected by clouds was tested and compared to the previously used version based on brightness temperature contrast. The neural network method decreased uncertainty between AERI and forward model spectra by 30 percent when compared with the previously used version.
First guess CO profiles to the AERI retrieval were from two different sources. One source was an a priori CO profile calculated as the mean profile from 56 individual measurements where each CO profile encompasses tower, aircraft, and satellite CO measurements. The other first guess CO profile came from the AIRS version 5 (AIRSv5) retrieved CO product. Incorporating the AIRS CO profile to the AERI retrieval provided a better estimate of free tropospheric CO when compared with the a priori profile. Using a better upper tropospheric CO estimate resulted in more accurate results from the AERI retrieval below 2 km, thus revealing that an AERI plus AIRS retrieved CO product is superior to either instrument's own CO retrieval working alone. The combined retrieval product is shown to have an RMSE of 10% in the first 2 km of the atmosphere.