Application of ANFIS model for prediction of zenith tropospheric delay

Tropospheric delay plays a crucial role in meteorological studies and weather forecasts as well as positioning accuracy. GPS derived zenith tropospheric delay (ZTD) has been established two decades ago. However, the main problem that still present is not all strategic points around the world will have a GPS receiver. To solve this shortcoming, a neural network with a fuzzy inference system that uses a back propagation algorithm is proposed to estimate ZTD value. The input of the system is surface meteorological data and the test output is ZTD from GPS. For a test case, a combination between surface pressure, temperature or relative humidity is performed to obtain the best estimation of ZTD. With data gathered over Antarctica: Casey, McMurdo and Syowa stations found that the ZTD estimated with three meteorological inputs (pressure, temperature and relative humidity, PTH) is agreed very well with ZTD from GPS than the input only from pressure (P), temperature (T), or pressure and temperature (PT), or pressure and relative humidity (PH). The accuracy of ZTD with PTH inputs is 20% higher with RMSE of 0.0175 mm than the other inputs. Thus, the ZTD estimated from surface meteorological data using ANFIS is beneficial and more useful when the GPS receiver at a specific location is absent.