Improving the performance of tropospheric mapping function in low elevation angle using artificial neural network

Abstract

Tropospheric delay is one of the major error sources in global navigation satellite systems (GNSS). The position accuracy determined using GNSS depends on the satellite’s geometry factor. However, in low elevation angle GNSS signals have a high tropospheric slant delay error and need a good mapping function model to cover this error. In this paper, two new tropospheric mapping functions (hydrostatic and non-hydrostatic) were developed based on an artificial neural network (ANN). The first model (Model A) was based on the day of the year (DOY), latitude and longitude. The second model (Model B) added real-time surface pressure to the parameters of Model “A.” The two developed models showed a strong correlation with the ray tracing mapping function, with a significant t-test value.The applicability of these mapping functions was tested over the Egypt region. The comparison was done with ray tracing of 5220 traces for 145 days from 2014 to 2017 as well as by assessing it (Mohamed et al., Vienna and Niell). This evaluation showed that the new model “B” agrees very well with the ray trace value. Model “B” and ray traces are on par with each other for a very low elevation angle 2⁰ the hydrostatic and non-hydrostatic mapping functions. Models “B” and “A” gave hydrostatic mean biases of 32 and 39 mm, respectively, and for non-hydrostatic, 4.5 and 4.9 mm, respectively. Vienna, Niell, and Mohamed et al. gave 46, 45, and 94 mm in hydrostatic, respectively, and 5.4, 5.2 and 9.2 mm in non-hydrostatic, respectively.