Comparative analysis of blind tropospheric correction models in Ghana

Abstract

Abstract The impact of the earth’s atmospheric layers, particularly the troposphere on Global Navigation satellite system (GNSS) signals has become a major concern in GNSS accurate positioning, navigation, surveillance and timing applications. For precise GNSS applications, tropospheric delay has to be mitigated as accurately as possible using tropospheric delay prediction models. However, the choice of a particular prediction model can signifi-cantly impair the positioning accuracy particularly when the model does not suit the user’s environment. A performance assessment of these prediction models for a suitable one is very important. In this paper, an assessment study of the performances of five blind tropospheric delay prediction models, the UNB3m, EGNOS, GTrop, GPT2w and GPT3 models was conducted in Ghana over six selected Continuously Operating Reference Stations (CORS) using the 1˚x1˚ gridded Vienna Mapping Function 3 (VMF3) zenith tropospheric delay (ZTD) product as a reference. The gridded VMF3-ZTD which is generated for every six hours on the 1˚x1˚ grids was bilinearly interpolated both space and time and transferred from the grid heights to the respective heights of the CORS locations. The results show that the GPT3 model performed better in estimating the ZTD with an overall mean (bias: 2.05 cm; RMS: 2.53 cm), followed by GPT2w model (bias: 2.32cm; RMS: 2.76cm) and GTrop model (bias: 2.41cm; 2.82cm). UNB3m model (bias: 6.23 cm; RMS: 6.43 cm) and EGNOS model (bias: 6.70 cm; RMS: 6.89 cm) performed poorly. A multiple comparison test (MCT) was further performed on the RMSE of each model to check if there is significant difference at 5% significant level. The results show that the GPT3, GPT2w and GTrop models are significantly indifferent at 5% significance level indicating that either of these models can be employed to mitigate the ZTD in the study area, nevertheless, the choice of GPT3 model will be more preferable.