The humidity retrievals using BP neural network algorithm based on mexihat or morlet wavelet function in clear-sky

The paper presents three algorithms to retrieve the humidity profiles from the observing datasets of FENGYUN-3A (FY-3A) satellite microwave humidity sensor. Compared to widely used algorithm-BP artificial neural network (BP-ANN), the algorithms of ANN based on morelet and mexihat function can avoid the conditions of easily getting stuck in local minima and reaching convergence with long time. Especially, the BP-ANN with mexihat function can decrease the training time and improve the humidity retrievals.