Skip to content

Estimating radar backscatter from snowpit observations, via DMRT and Machine Learning

License

Notifications You must be signed in to change notification settings

snowex-hackweek/radar_prediction

Repository files navigation

Radar Scattering Prediction

Project Summary

We are trying to estimate the backscattering of radar from snowpits using parameters from SnowEx data such as density, temperature, grain size and height. We are also trying to understand the relationship between these input parameters and the backscatter and evaluate the parameter sensitivity. We also want to explore the link between SnowEx pit measurements and input parameters of DMRT-Bic. Another cool thing would be to see if there is a correlation between the SWE and the amplitudes.

The problem

Currently, we know that there is backscatter when a radar is directed towards snow but we do not know what variables of the snow actually have an impact on this backscatter and how much impact each of these parameters have on it. This is is the problem we are trying to solve using Machine Learning. In the broader context, the inference of the relationship between input parameters and the amplitude can help us calibrate current sensor tools and also to be used to reverse engineer when the satellites are up in the air.

Application Example

Decsision Tree Model (for VV polarization):

Decision Tree

Random Forest Model (for VV polarization):

Random Forest

Variable Importance (for VV polarization):

Variable Importance

Sample data

Snow pit data set accessed from the snowexsql database:

snowex

UAVSAR Data for Output:

uavsar

Specific Questions

  1. Can we use machine learning to predict the radar backscattering from snowex input parameters?
  2. Is there a relationship between the DMRT input parameters and SnowEx pit measurements?
  3. How much do the input parameters contribute to the variability of the amplitude?
  4. Is there a correlation between the SWE and the amplitudes?

Existing methods

The existing method to do this is using a DMRT Model.

Proposed methods/tools

We are implementing linear regression, decision trees and random forest machine learning models to helps us answer the questions.

Collaborators on this project

Team Member Role GitHub ID
Jonas Jans Project Lead jonas-frederik
HP Marshall Data Science Lead hpmarshall
Aaliyah Hanni Data Engineer aaliyahfiala42
Yiying Gao ML Engineer viggieG
Anuhya Bhagavatula Data Scientist anuhyabs
Shrusti Ghela ML Engineer shrusti-ghela
Samuel Marcus Intern samuelmarcus99

Background reading

  1. https://ieeexplore.ieee.org/document/6185696
  2. https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=6877612
  3. https://web.eecs.umich.edu/~leutsang/Computer%20Codes%20and%20Simulations.html
  4. https://ntrs.nasa.gov/api/citations/20150000366/downloads/20150000366.pdf
  5. https://snowexsql.readthedocs.io/en/latest/gallery/raster_union_and_more_example.html

About

Estimating radar backscatter from snowpit observations, via DMRT and Machine Learning

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published