The fitting of a polynomial model to calculate an orbital error estimate could result in overfitting the data if a model other then linear is chosen, particularly for large-scale Sentinel-1 data which come with relatively accurate orbit data compared to older sensors such as ERS or Envisat.
Todo: conduct intensive tests using different data sets and sensors on the orbital error estimate when using different settings for the polynomial degree (and potentially also comparing the independent and network methods). The attached bash script (saved as .txt) can be used to perform automated PyRate runs of the 'correct' step using different parameter settings. Based on the outcome, recommendations on the polynomial degree can be made in the PyRate documentation or even warning messages added to the logs when using a polynomial degree not recommended for a particular sensor.
The fitting of a polynomial model to calculate an orbital error estimate could result in overfitting the data if a model other then linear is chosen, particularly for large-scale Sentinel-1 data which come with relatively accurate orbit data compared to older sensors such as ERS or Envisat.
Todo: conduct intensive tests using different data sets and sensors on the orbital error estimate when using different settings for the polynomial degree (and potentially also comparing the independent and network methods). The attached bash script (saved as .txt) can be used to perform automated PyRate runs of the 'correct' step using different parameter settings. Based on the outcome, recommendations on the polynomial degree can be made in the PyRate documentation or even warning messages added to the logs when using a polynomial degree not recommended for a particular sensor.
run_pyrate_orbital.txt