Masters Theses

Keywords and Phrases

GNSS; InSAR; Machine Learning; Tropospheric; ZTD


"Interferometric Synthetic Aperture Radar (InSAR) is a popular technique for studying Earth's surface deformation caused by phenomena like earthquakes and subsidence. However, its accuracy is limited by tropospheric delays caused by water vapor in the atmosphere. This limitation can be overcome by using methods that correct for tropospheric noise, such as statistical, empirical, and predictive approaches. This study explores the potential of using machine learning algorithms to predict the zenith total delay caused by tropospheric effects in InSAR measurements. The study employs two different machine learning algorithms, random forest and neural networks, to learn the relationship between numerical weather prediction model data and InSAR parameters in Continental USA and the globe. The neural network model outperforms both the random forest model and the traditional approach, reducing the RMSE by approximately 30%. The study demonstrates that machine learning algorithms can effectively correct tropospheric noise in most interferograms, resulting in a 30-60% improvement in Pennsylvania and Hawaii. However, the neural network model faces challenges in making predictions in areas with high variability in local climate and weather patterns. Overall, this research presents a promising approach for improving InSAR accuracy by using machine learning algorithms to correct for tropospheric noise"--Abstract, p. iii


Maurer, Jeremy

Committee Member(s)

Smith, Ryan G.
Tripathy, Ardhendu S.


Geosciences and Geological and Petroleum Engineering

Degree Name

M.S. in Geological Engineering


Missouri University of Science and Technology

Publication Date

Spring 2023


xi, 80 pages

Note about bibliography

Includes_bibliographical_references_(pages 75-79)


© 2023 Ngo Hi Kenny Yue, All Rights Reserved

Document Type

Thesis - Open Access

File Type




Thesis Number

T 12270

Electronic OCLC #