Exploring SAR interferometry (InSAR) for snow water equivalent (SWE) estimation

Exploring SAR interferometry (InSAR) for snow water equivalent (SWE) estimation

Requirements

  • M.Sc. in Machine Learning, Data Science, Computer Science, Mathematics, Telecommunications, or similar
  • Knowledge of Python
  • Software development skills
  • Basic concepts of signals processing
  • Basic concepts of data science, concerning data analysis, processing and machine learning

Description

Snow Water Equivalent (SWE) is a crucial variable for understanding water resources and climate dynamics, particularly in regions where snowmelt is a primary source of freshwater. Traditional methods for estimating SWE often rely on ground-based measurements, which are spatially limited, or passive satellite sensors, which struggle to penetrate thick snowpacks. Synthetic Aperture Radar (SAR) Interferometry (InSAR) presents a promising remote sensing technique to address these limitations by offering high-resolution, all-weather imaging capabilities, particularly in snow-covered terrains. However, accurately extracting SWE information from SAR signals remains a challenge due to the complex interaction between radar waves and snow properties.

This research proposes leveraging the potential of InSAR for SWE estimation by incorporating state-of-the-art deep learning models. By exploiting the phase differences in multi-temporal SAR images, InSAR can provide insights into changes in snow volume and density. The primary aim of this study is to develop a robust deep learning framework capable of mapping InSAR-derived features to SWE estimates. The proposed approach will combine both traditional SAR processing techniques with deep learning architectures, such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), to capture both spatial and temporal dynamics of snow cover.

The research will explore several key objectives:

  • Literature review of dataset and models
  • Development and evaluation of deep learning models for SWE estimation trained on datasets found during literature review.
  • Comparative analysis of the proposed model’s performance

This project is expected to contribute to improving water resource management and hydrological modeling by offering more accurate and scalable SWE estimates. The combination of InSAR and deep learning represents a novel approach that could enhance our understanding of snow dynamics, supporting climate monitoring and water management efforts globally.

Contacts