Modeling of zenith path delay over Antarctica using an adaptive neuro fuzzy inference system technique

Accessibility and accurate estimation of the tropospheric delay plays a crucial role in meteorological studies and weather forecasts as well as improving positioning accuracy. We propose to employ an adaptive neuro fuzzy inference system (ANFIS) to build estimation and prediction models for zenith path delay (ZPD). Five selected stations over Antarctica were used to examine the applicability of ANFIS. GPS ZPD data of 2010 with five-minute resolution was used as the target output. A fuzzy clustering algorithm is adopted to enhance the performance of the models, which is able to minimize the number of membership functions and rules for better efficiency in the models. To investigate the accuracy of models developed, a combination of the surface pressure (P), temperature (T) and relative humidity (H) is performed to obtain the best estimation of ZPD. The results demonstrated that ANFIS models with three inputs network (P, T and H) agreed very well with ZPD obtained from GPS than separated input only coming from P or T, or P and T, or P and H. Finally, the input network (P, T and H) is selected in developing the ZPD predictive models. The prediction resulted from one-step to eight-step ahead development, demonstrated that the high-resolution of data used in training process will increase the accuracy of the predictive model.