Web Analytics

TropDS: Downscaling to Enhance Tropospheric Delay Grid Precision in Space Geodesy

ZHD blur ZHD clear
RMSE ZHD before RMSE ZHD after
ZWD blur ZWD clear
RMSE ZWD before RMSE ZWD after
ZWD blur ZWD clear
RMSE ZWD before RMSE ZWD after

Abstract

We introduce TropDS, a downscaling framework that combines physical residual modeling with a U-Net AI refinement to produce high-resolution tropospheric delays from coarse global grids. Global ZHD & ZWD accuracies reach 90.97% & 85.76%, with RMSE < 2 mm and inference taking ~0.04 s per epoch. It also remarkably boosts early-stage forecast precision when coupled with AI weather models like Pangu-Weather, GraphCast and FengWu. A robust solution for precise atmospheric corrections in geodesy.

BibTeX

@article{ding2026tropds, title={TropDS: Downscaling to Enhance Tropospheric Delay Grid Precision in Space Geodesy}, author={Ding, Junsheng et al}, journal={Journal of Geodesy}, year={2026}, }

Total Visits: