Author not pictured:
Mohamed Tamazin Formerly a GNSS Engineering Manager at Safran Electronics & Defense

This white paper presents a hybrid modeling framework for tropospheric delay in GNSS signal simulation, combining empirical formulations with a machine learning-based prediction of the Zenith Wet Delay (ZWD).

While conventional models accurately estimate the hydrostatic component, they remain limited in capturing the high spatial-temporal variability of the wet component. To address this limitation, a feedforward neural network trained on over two decades of global GNSS observations and associated meteorological data is employed to improve ZWD estimation.

The proposed approach demonstrates a significant reduction in modeling errors and provides a more realistic representation of atmospheric effects, particularly under dynamic weather conditions.