Improved ANN method for retrieving Asian dust AOT and altitude from AIRS measurements

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 7 January 2015
Byeong-Gwon Kim, Seoul National University, Seoul, South Korea; and H. Han and B. J. Sohn

Retrieval algorithms for aerosol optical thickness (AOT) and altitude of the mineral dust from Atmospheric Infrared Sounder (AIRS) measurements were developed by Han and Sohn (2013) using artificial neural network (ANN) approach. It has been noted that the algorithm has difficulty in retrieving for the long-term data processing because of degrading channels and unknown information such as cloud screening. In order to improve the performance of those dust retrieval algorithms, new training datasets were constructed based on new channels selection with sensitivity test and data quality check. Now new ANN model was trained by relating AIRS brightness temperatures, surface elevation, and relative air mass to collocated AOT from the Moderate Resolution Imaging Spectroradiometer (MODIS) and mean dust height derived from Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) if Asian dusts are detected. The improved ANN model shows better result in spite of smaller number of used input channels, compared with Han and Sohn (2013). For the time being we plan to build up dust climatology for over 10 years, from which climate change related signals can be detected, about dust outbreak changes in response to the global warming.