2026 Undergraduate Research Showcase

Snow Depth Estimation via Machine Learning on L-Band InSAR Backscatter and Lidar Data

Document Type

Student Presentation

Presentation Date

4-24-2026

Faculty Sponsor

Dr. Hans-Peter Marshal and Dr. Ibrahim Alabi

Abstract

The US heavily relies on snowmelt to meet water supply demands. 60-75% of all fresh water in the western US comes from snowmelt, and 50% of global freshwater is derived from snow. Currently, there is no practical way to measure snow depth across large and remote regions, making it critical to develop methods for estimating snow accumulation at scale. NASA’s NISAR mission, which launched in July 2025, will provide global L-band SAR data every 12 days. This study explores the use of various machine learning models to predict snow depth by training on InSAR data collected in 2021 with LiDAR measurements as ground truth. We surveyed 3 sites, Banner Summit (167.1 km²), Mores Creek (34.7 km²), and Grand Mesa(62.7 km²). By combining multiple regions with varying terrain and environmental conditions, the dataset reduces the risk of overfitting to a single location and improves the model’s ability to generalize to real-world applications. The models achieved snow depth predictions within approximately 10 cm of ground truth measurements across all sites. This level of accuracy demonstrates the feasibility of leveraging the NASA NISAR mission for scalable, global snow monitoring, with significant implications for water resource management and climate research.

This document is currently not available here.

Share

COinS