A neural network based algorithm for the retrieval of TPW from AMSU measurements

Total Precipitable Water (TPW) in a column of atmosphere is one of the important parameters useful for a number of meteorological applications. In the present study, a neural network based algorithm has been developed for the retrieval of TPW using NOAA-16 AMSU measurements. The TPW has been derived experimentally using NOAA-16 AMSU measurements locally received from High Resolution Picture Transmission (HRPT) station at India Meteorological Department (IMD) separately over ocean only. The validation of TPW has been carried out against the TPW derived from Radiosonde (RAOB) data. The bias and rms errors against the RAOB derived TPW have been found to about 0.11 mm and 2.98 mm respectively. The intercomparisions of TPW derived using NOAA AMSU data have also been made with that of NOAA/NESDIS derived TPW. Further, case study for the potential use of TPW derived from NOAA AMSU data has been carried out. This case study has revealed that the concentration of maximum precipitable water values in conjunction with high Sea surface wind speed data from Quickscat Scatterometer were found very useful for forecasting the heavy to very heavy rainfall event along the west coast of India. Therefore, AMSU derived TPW could be used as an important parameter for the operational weather forecasting on a real time basis.