InSAR Displacement Calibration using GPS Data

Overview

Project for UTD's PHYS5336 course with the following objectives:

• Supervised classification of InSAR data points that are affected by noises and decorrelation, and thus needs calibration.
• Calibration of InSAR data points using supervised regression.

For the sake of the project, we make the following assumption:

• GPS data is always correct (which is not always true).

Data

• SAR data is obtained from European Space Agency's Sentinel-1 satellites.
• The SAR data is processed using ISCE (InSAR Scientific Computing Environment) and MintPy.
• We use GPS data from EarthScope Consortium's Network of the Americas (NOTA) GNSS stations.

Files

Classification.ipynb

Performs supervised classification of InSAR data points that are affected by noises and decorrelation, and thus needs calibration.

Regression.ipynb

Predicts calibration factor for calibration of InSAR data points using supervised regression.

Input Data Format

The input data (inputs-for-ml/final_ml_data.csv) should be of the following format:

date	station	latitude	longitude	elevation	slope	north_gps	east_gps	vertical_gps	coherence	los_insar	bias	needs_calibration
YYYY-MM-DD	NAME	lat	lon	elev	slope	$d_N$	$d_E$	$d_V$	0 - 1	$d_{LOS}$	0.0	0/1