Deep Neural Networks for Refining Vertical Modeling of Global Tropospheric Delay

Research output: Contribution to journal › Article › Research › peer review

Authors

  • Peng Yuan
  • Kyriakos Balidakis
  • Jungang Wang
  • Pengfei Xia
  • Jian Wang
  • Mingyuan Zhang
  • Weiping Jiang
  • Harald Schuh
  • Jens Wickert
  • Zhiguo Deng

External Research Organisations

  • Helmholtz Centre Potsdam - German Research Centre for Geosciences (GFZ)
  • Technische Universität Berlin
  • Wuhan University
View graph of relations

Details

Original languageEnglish
Article numbere2024GL111404
JournalGeophysical research letters
Volume52
Issue number2
Publication statusPublished - 25 Jan 2025
Externally publishedYes

Abstract

Kinematic airborne platforms are becoming increasingly vital for Earth observation. They highlight the critical need for accurate tropospheric delay corrections across varying altitudes, especially as most existing models are limited to Earth's surface. Although analytical functions have been used to model vertical reductions in tropospheric delays, they struggle to capture the intricate vertical variations of atmospheric state. In response, we introduce a novel approach that utilizes deep neural networks (DNN) to reconstruct global three-dimensional zenith hydrostatic delay (ZHD) and zenith wet delays (ZWD) derived from numerical weather models (NWM). Our method reconstructs NWM-derived ZHD and ZWD globally up to 14 km above the Earth's surface, with average precision levels of 0.4 and 0.8 mm, respectively. Compared to the analytical third-order exponential model, the DNN approach demonstrates substantial improvement with global average root-mean-square reductions of 63% for ZHD and 36% for ZWD.

Keywords

    deep neural networks, ERA5, GNSS, GNSS meteorology, tropospheric delay, vertical correction model

ASJC Scopus subject areas

Cite this

Deep Neural Networks for Refining Vertical Modeling of Global Tropospheric Delay. / Yuan, Peng; Balidakis, Kyriakos; Wang, Jungang et al.
In: Geophysical research letters, Vol. 52, No. 2, e2024GL111404, 25.01.2025.

Research output: Contribution to journal › Article › Research › peer review

Yuan, P, Balidakis, K, Wang, J, Xia, P, Wang, J, Zhang, M, Jiang, W, Schuh, H, Wickert, J & Deng, Z 2025, 'Deep Neural Networks for Refining Vertical Modeling of Global Tropospheric Delay', Geophysical research letters, vol. 52, no. 2, e2024GL111404. https://doi.org/10.1029/2024GL111404
Yuan, P., Balidakis, K., Wang, J., Xia, P., Wang, J., Zhang, M., Jiang, W., Schuh, H., Wickert, J., & Deng, Z. (2025). Deep Neural Networks for Refining Vertical Modeling of Global Tropospheric Delay. Geophysical research letters, 52(2), Article e2024GL111404. https://doi.org/10.1029/2024GL111404
Yuan P, Balidakis K, Wang J, Xia P, Wang J, Zhang M et al. Deep Neural Networks for Refining Vertical Modeling of Global Tropospheric Delay. Geophysical research letters. 2025 Jan 25;52(2):e2024GL111404. doi: 10.1029/2024GL111404
Yuan, Peng ; Balidakis, Kyriakos ; Wang, Jungang et al. / Deep Neural Networks for Refining Vertical Modeling of Global Tropospheric Delay. In: Geophysical research letters. 2025 ; Vol. 52, No. 2.
Download
@article{91b3a9ac9adc40aa9950f40da996d200,
title = "Deep Neural Networks for Refining Vertical Modeling of Global Tropospheric Delay",
abstract = "Kinematic airborne platforms are becoming increasingly vital for Earth observation. They highlight the critical need for accurate tropospheric delay corrections across varying altitudes, especially as most existing models are limited to Earth's surface. Although analytical functions have been used to model vertical reductions in tropospheric delays, they struggle to capture the intricate vertical variations of atmospheric state. In response, we introduce a novel approach that utilizes deep neural networks (DNN) to reconstruct global three-dimensional zenith hydrostatic delay (ZHD) and zenith wet delays (ZWD) derived from numerical weather models (NWM). Our method reconstructs NWM-derived ZHD and ZWD globally up to 14 km above the Earth's surface, with average precision levels of 0.4 and 0.8 mm, respectively. Compared to the analytical third-order exponential model, the DNN approach demonstrates substantial improvement with global average root-mean-square reductions of 63% for ZHD and 36% for ZWD.",
keywords = "deep neural networks, ERA5, GNSS, GNSS meteorology, tropospheric delay, vertical correction model",
author = "Peng Yuan and Kyriakos Balidakis and Jungang Wang and Pengfei Xia and Jian Wang and Mingyuan Zhang and Weiping Jiang and Harald Schuh and Jens Wickert and Zhiguo Deng",
note = "Publisher Copyright: {\textcopyright} 2025. The Author(s).",
year = "2025",
month = jan,
day = "25",
doi = "10.1029/2024GL111404",
language = "English",
volume = "52",
journal = "Geophysical research letters",
issn = "0094-8276",
publisher = "Wiley-Blackwell",
number = "2",

}

Download

TY - JOUR

T1 - Deep Neural Networks for Refining Vertical Modeling of Global Tropospheric Delay

AU - Yuan, Peng

AU - Balidakis, Kyriakos

AU - Wang, Jungang

AU - Xia, Pengfei

AU - Wang, Jian

AU - Zhang, Mingyuan

AU - Jiang, Weiping

AU - Schuh, Harald

AU - Wickert, Jens

AU - Deng, Zhiguo

N1 - Publisher Copyright: © 2025. The Author(s).

PY - 2025/1/25

Y1 - 2025/1/25

N2 - Kinematic airborne platforms are becoming increasingly vital for Earth observation. They highlight the critical need for accurate tropospheric delay corrections across varying altitudes, especially as most existing models are limited to Earth's surface. Although analytical functions have been used to model vertical reductions in tropospheric delays, they struggle to capture the intricate vertical variations of atmospheric state. In response, we introduce a novel approach that utilizes deep neural networks (DNN) to reconstruct global three-dimensional zenith hydrostatic delay (ZHD) and zenith wet delays (ZWD) derived from numerical weather models (NWM). Our method reconstructs NWM-derived ZHD and ZWD globally up to 14 km above the Earth's surface, with average precision levels of 0.4 and 0.8 mm, respectively. Compared to the analytical third-order exponential model, the DNN approach demonstrates substantial improvement with global average root-mean-square reductions of 63% for ZHD and 36% for ZWD.

AB - Kinematic airborne platforms are becoming increasingly vital for Earth observation. They highlight the critical need for accurate tropospheric delay corrections across varying altitudes, especially as most existing models are limited to Earth's surface. Although analytical functions have been used to model vertical reductions in tropospheric delays, they struggle to capture the intricate vertical variations of atmospheric state. In response, we introduce a novel approach that utilizes deep neural networks (DNN) to reconstruct global three-dimensional zenith hydrostatic delay (ZHD) and zenith wet delays (ZWD) derived from numerical weather models (NWM). Our method reconstructs NWM-derived ZHD and ZWD globally up to 14 km above the Earth's surface, with average precision levels of 0.4 and 0.8 mm, respectively. Compared to the analytical third-order exponential model, the DNN approach demonstrates substantial improvement with global average root-mean-square reductions of 63% for ZHD and 36% for ZWD.

KW - deep neural networks

KW - ERA5

KW - GNSS

KW - GNSS meteorology

KW - tropospheric delay

KW - vertical correction model

UR - http://www.scopus.com/inward/record.url?scp=85216215167&partnerID=8YFLogxK

U2 - 10.1029/2024GL111404

DO - 10.1029/2024GL111404

M3 - Article

AN - SCOPUS:85216215167

VL - 52

JO - Geophysical research letters

JF - Geophysical research letters

SN - 0094-8276

IS - 2

M1 - e2024GL111404

ER -