Snow depth estimation using SAR tomography (Develop a machine learning tool to discover Under glacier craters)

M-GEO

ACQUAL

Staff Involved

M-SE Core knowledge areas

Spatial Information Science (SIS)

Topic description

As global temperatures rise, the melting of the world's snow resources is causing sea levels to rise at an ever-increasing rate. Knowledge of these resources and snow water equivalent plays an important roles for ecosystem, human use of snow resources, prediction of economic losses, etc. In this context, snow depth is one of the most important parameters that primarily quantifies snow resources. This important metric cannot be measured directly with satellite imagery because the information is lost in the vertical dimension due to 2D imaging system. Recently, advanced SAR tomography enabled the generation of a full 3D image and the recovery of the lost vertical information. This advanced technology provides the ability to accurately estimate snow depth using penetrating SAR images and discover massive craters under snow and glaciers. An example of the recent application of this technology was reported by NASA on the discovered a crater under the Greenland ice sheet, which is much larger than Washington DC and Paris (see the figure and link below).
https://scitechdaily.com/nasa-discovers-a-massive-crater-under-greenland-ice/

Topic objectives and methodology

In this project, the student will develop a machine learning method to reconstruct a vertical image and estimate snow depth. Available polarimetric SAR data from airborne acquisition will be used for this project. The idea is to enhance the ability of SAR tomography to a generate 3D image using a machine learning method. The main goal is to develop a model that recognizes the vertical structure of the snow and inspects the ground surface hidden below snow and particularly identifies holes under the snow at great depth.

References for further reading

https://scitechdaily.com/nasa-discovers-a-massive-crater-under-greenland-ice/