A multiple linear regression method and a neural network method are performed to retrieve the Precipitable Water Vapor, surface temperature and relative humidity using microwave (AMSU-A, MHS) and infrared (HIRS) ATOVS sounders. Each method is performed using microwave, infrared, and mixed data separately to assess the best. Near nadir ATOVS data of Dakar region (Senegal) at 12:00 AM and 12:00 PM are used for the whole year 2013. Learning data are from radiosonde and in situ measurements. By comparing them with retrieved data, ECMWF reanalysis data help to validate the different methods. The multiple linear regression method provides good results for microwave data with an RMS of 4.65 mm, 2.27