ETH ZURICH / SPACE GEODESY / GNSS / AI

About Me

I am Zhenyi Zhang, a PhD candidate in the Space Geodesy Group at ETH Zurich, supervised by Prof. Benedikt Soja.

My research builds on a strong background in space geodesy, atmospheric science, and machine learning. My doctoral work focuses on physics-informed machine learning for tropospheric modeling, integrating geodetic observations, meteorological data, and physical knowledge. I developed the first global AI-based models for asymmetric tropospheric mapping functions and direct slant wet delay prediction.

I also contribute to the Earth System Foundation Model (ESFM) project and the Swiss AI Initiative. My contribution to ESFM focuses on integrating satellite water vapor observations into the model. More broadly, I explore physics-informed frameworks that combine multi-source satellite observations and meteorological data for atmospheric modeling and prediction, with applications in weather forecasting.

Previously, I received my MSc and BSc from Wuhan University. During my master's studies, I was fortunate to be mentored by Prof. Yidong Lou and Prof. Weixing Zhang. My research focused on high-precision processing of ground-based and highly dynamic airborne GNSS observations, as well as the generation of GNSS-derived tropospheric products. This work supported applications in data assimilation, numerical weather prediction, and InSAR atmospheric correction.

Email: zhenyzhang@ethz.ch

Research Interests

• AI for Science
  Physics-informed machine learning and foundation models for atmospheric and Earth-system science, with applications in water vapor forecasting and extreme weather detection.
• Specialized AI models for meteorology and space geodesy
  Development of specialized machine learning models that incorporate physical constraints for applications in meteorology and space geodesy, including tropospheric mapping functions and direct slant wet delay prediction.
• GNSS data processing & meteorology
  Precise processing of ground-based and highly dynamic airborne GNSS observations, tropospheric estimation, and atmospheric water vapor retrieval, with applications in GNSS data assimilation for numerical weather prediction and InSAR atmospheric correction.
• Geodetic observations for climate research
  Applications of geodetic observations, particularly GNSS-derived tropospheric products, to climate monitoring and prediction, including ENSO prediction and changes in the global distribution and vertical structure of atmospheric water vapor.

Selected Publications

See Google Scholar for the full publication list. † Corresponding author.

2026 · Accepted

Assessing the Impact of Dynamic Tropospheric Stochastic Constraints on Airborne GNSS Positioning and ZTD Estimation

Z. Zhang, Y. Lou, W. Zhang^†, Y. Ouyang, Y. Zhou, J. Bai, B. Soja

Measurement

Evaluates dynamic tropospheric constraints in airborne GNSS experiments, examining their effects on positioning, zenith delays, and coupled estimation parameters.

Paper

2026 · Published

Machine-learned mapping functions for the asymmetric troposphere

Z. Zhang^†, B. Soja

Journal of Geodesy, 100, 79

Global asymmetric tropospheric mapping functions learned through physics-constrained neural networks, using ERA5 ray tracing to represent direction-dependent atmospheric delays.

PaperCode

2026 · Published

MSWD: A hybrid machine-learning framework for slant wet delay modeling

Z. Zhang^†, B. Soja

Journal of Geophysical Research: Atmospheres

A hybrid neural model embeds a physical tilting formulation to predict slant wet delays directly, with optional temperature and humidity inputs.

PaperCode

2026 · Preprint

Earth System Foundation Model (ESFM): A unified framework for heterogeneous data integration and forecasting

F. Ozdemir^†, Y. Cheng, S. Mohebi, F. Lehmann, S. Adamov, Z. Zhang, L. Trentini, D. Grund, O. Fuhrer, T. Hoefler, S. Mishra, S. Schemm, B. Soja, M. Salzmann

Communications Earth & Environment - Under revision (preprint)

An open Earth-system foundation model combines gridded, satellite, and station observations within one backbone for flexible forecasting and downstream adaptation.

arXivCodeProjectPDF

2025 · Published

Assimilating ground-based and high-dynamic airborne GNSS zenith total delays into numerical weather predictions

Z. Zhang^†, M. Liu, W. Zhang^†, Y. Lou, B. Soja

IEEE Transactions on Geoscience and Remote Sensing

Combines ground-based and UAV-derived GNSS zenith delays in WRF data assimilation to improve the atmospheric humidity field.

Paper

2023 · Published

Dynamic stochastic model for estimating GNSS tropospheric delays from air-borne platforms

Z. Zhang, Y. Lou, W. Zhang^†, Z. Wang, Y. Zhou, J. Bai, Z. Zhang, C. Shi

GPS Solutions

Adapts tropospheric process noise to spatial and temporal atmospheric variability, improving high-dynamic airborne GNSS zenith-delay estimation.

Paper

2022 · Published

Assessment of ERA-Interim and ERA5 reanalysis data on atmospheric corrections for InSAR

Z. Zhang, Y. Lou, W. Zhang^†, H. Wang, Y. Zhou, J. Bai

International Journal of Applied Earth Observation and Geoinformation

Compares ERA5 and ERA-Interim for InSAR atmospheric correction across seasons and climate regimes, highlighting limitations in resolving small-scale turbulence.

Paper

2021 · Published

On the Assessment GPS-Based WRFDA for InSAR Atmospheric Correction: A Case Study in Pearl River Delta Region of China

Z. Zhang, Y. Lou, W. Zhang^†, H. Wang, Y. Zhou, J. Bai

Remote Sensing

Assesses GPS zenith-delay assimilation in WRFDA for correcting InSAR atmospheric errors in China's Pearl River Delta.

Paper

Academic Service

Peer Review

Journal of Geodesy; Journal of Advances in Modeling Earth Systems; IEEE Transactions on Geoscience and Remote Sensing; Atmospheric Research; Earth and Space Science; Measurement; Advances in Space Research; IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.

Contact

Space Geodesy Group
Institute of Geodesy and Photogrammetry
ETH Zurich, Switzerland

zhenyzhang@ethz.ch