Systematic Dry Bias and Geographic Dependencies in a High-Resolution NWM's Zenith Total Delay Revealed by GNSS and Radiosonde Validation

Systematic Dry Bias and Geographic Dependencies in a High-Resolution NWM's Zenith Total Delay Revealed by GNSS and Radiosonde Validation

 

¹Tsebeje, S. Y., ^1,2Wang, J., ³Dodo, J. D. and ^1,2Schuh, H.

           

1) Technische Universität Berlin, Berlin, Germany

2) GFZ, Helmholtz Centre for Geosciences, Potsdam, Germany.

3) Centre for Geodesy and Geodynamics (CGG) National Space Research and Development     

     Agency (NASRDA), Toro. Nigeria.

 

 

Abstract

This study reveals a systematic dry bias and distinct geographic dependencies in high-resolution Numerical Weather Model ERA5 (NWM) Zenith Total Delay (ZTD) estimates, as comprehensively validated against GNSS and Radiosonde (RS) observations for 2022. We analyzed data from 13 African stations, including four collocated sites with RS and GNSS reference points. While the NWM shows excellent agreement with RS data (mean RMSE: 0.0009 m, R > 0.996), a consistent dry bias is evident when compared with the GNSS-derived ZTD, averaging –0.0042 m at the collocated sites. The bias is moderately correlated with station elevation (R = –0.731), indicating a poorer model performance at higher altitudes. Importantly, spatial interpolation from the NWM grid to non-collocated GNSS sites did not introduce a statistically significant additional bias (p-value: 0.7719), indicating that the error was intrinsic to the model rather than its post-processing. Furthermore, a significant temporal error autocorrelation and large dry bias in the Integrated Water Vapour were identified. The findings highlight the model's water vapour parameterization, especially over complex terrain, as the primary source of error rather than spatial representativeness, with clear evidence for prioritizing improvements in the physical formulation of the model over adjustments to interpolation strategies.