Estimation of precipitable water vapor using an adaptive neuro-fuzzy inference system technique

Abstract

Water vapor has an important role in the global climate change development. Because it is essential to human life, many researchers proposed the estimation of atmospheric water vapor values such as for meteorological applications. Lacking of water vapor data in a certain area will a problem in the prediction of current climate change. Here, we reported a novel precipitable water vapor (PWV) estimation using an adaptive neuro-fuzzy inference system (ANFIS) model that has powerful accuracy and higher level. Observation of the surface temperature, barometric pressure and relative humidity from 4 to 10 April 2011 has been used as training and the PWV derived from GPS as a testing of these models. The results showed that the model has demonstrated its ability to learn well in events that are trained to recognize. It has been found a good skill in estimating the PWV value, where strongest correlation was observed for UMSK station (r = 0.95) and the modest correlation was for NTUS station (r = 0.73). In general, the resulting error is very small (less than 5%). Thus, this model approach can be proposed as an alternative method in estimating the value of PWV for the location where the GPS data is inaccessible. © 2013 Springer-Verlag.