Development of a neural network algorithm for the retrieval of TPW from NOAA16 AMSU measurements

Abstract

A neural‐network‐based algorithm for the retrieval of Total Precipitable Water (TPW) using Advanced Microwave Sounding Unit (AMSU) data available in real time from NOAA16 satellite has been developed. The retrieval method benefits from reliable surface observations, which includes the skin temperature and ocean surface wind speed and direction. The algorithm uses the simulated brightness temperatures at four frequencies, 23.4 Ghz, 31.4 Ghz, 50.3 Ghz, and 89.0 Ghz, of AMSU‐A as input and TPW derived from Radiosonde Observations (RAOB) profiles as output. The pairs of input and output are restricted to the homogeneous emitting areas, e.g. over Bay of Bengal and Arabian Sea. The performance of the algorithm is assessed using independent RAOB measurements. The bias and rms differences are found to be 0.21 mm and 2.03 mm respectively over the range of 15 and 75 mm. Further, extensive comparisons are made between the TPW obtained from the neural network algorithm and those obtained using other satellite instruments like the Tropical Rainfall Measuring Mission Microwave Imager, Advanced Infrared Sounder, and Moderate Resolution Imaging Spectroradiometer, in which the results are found to be in close agreement.

Acknowledgements

The authors are grateful to India Meteorological Department for providing data. The authors are grateful to Indian Institute of Science, Bangalore for providing the radiosonde data. We are also thankful for Dr Mitch Goldberg NOAA/NESDIS/ORA for providing the Limb correction coefficients for NOAA‐16 AMSU. We are thankful for Dr N. C. Grody NOAA/NESDIS for his valuable input during the course of this study. The data used in this study were downloaded from the NOAA and NASA websites. All contour plots in this paper have been plotted using GrADS software.