Xihan Yao

Xihan Yao

Ph.D. Student
UT Austin

[Award] Second Place Award at UT Austin Geoscience Hackathon 2025

1 minute read

Published:

Excited to share that my team won Second Place at the 2025 UT Austin Geoscience Hackathon!

Project: Tracking Earth’s Changes with AlphaEarth Foundations

Team Members: Xihan Yao (myself), Rushi Bhatt, and mentor Dr. Brendon Hall

Project Theme: Change over time in Earth’s systems

What We Built

Our team developed an innovative open-source geospatial AI platform that leverages Google Earth Engine and DeepMind’s AlphaEarth embeddings to detect and visualize land-cover changes at planetary scale.

Technical Approach

We utilized AlphaEarth embeddings — a 64-dimensional representation trained on satellite imagery from 2017-2024 — to capture complex patterns in Earth’s surface. By computing dot products between embeddings from different years and applying threshold-based analysis, we can quantify changes across three dimensions:

  1. Temporal: When did the change occur?
  2. Magnitude: How significant is the change?
  3. Resolution: At what spatial scale can we detect the change?

Real-World Applications

Our platform successfully detects and visualizes diverse environmental changes, including:

  • Urban development: Construction and demolition of buildings on UT Austin campus
  • Wildfire impacts: Forest fire scars across California over multiple years
  • Urban sprawl: Expansion patterns in Austin, Texas
  • Agricultural rotation: Crop cycling patterns in Central Texas

Interactive Tool

We developed an interactive dashboard that enables users to:

  • Define custom areas of interest
  • Compute land-cover changes within seconds
  • Visualize results with intuitive visualizations
  • Export data for further analysis

This tool makes planetary-scale change detection accessible to researchers, policymakers, and environmental professionals.

Impact

This hackathon project demonstrates how cutting-edge AI embeddings and cloud-based geospatial processing can unlock new insights into Earth’s changing systems. We’re excited to continue developing this tool to make it available as a publicly accessible online platform that empowers decision-makers globally.

The combination of state-of-the-art AI models with interactive visualization creates powerful opportunities for environmental monitoring, climate adaptation planning, and sustainable resource management.

Repository: GitHub - AlphaEarth Foundations

The Geoscience Hackathon is an annual event organized by the Jackson School of Geosciences at UT Austin, supported by the UT Open Source Program Office (OSPO).

You May Also Enjoy

Connecting the Dots: A Geospatial Journey Through Place and Space

1 minute read

Published:

As part of my final project for the Issues in Geography class, I created an interactive globe visualization that traces my personal and academic journey across different places. This project, titled “Connecting the Dots,” explores the relationship between place and space through my own experiences.

[Conference] ACSP 2025 Conference in Minneapolis

3 minute read

Published:

I had the wonderful opportunity to attend and present at the Association of Collegiate Schools of Planning (ACSP) 2025 Annual Conference in Minneapolis, Minnesota, from October 25-27, 2025. This was my first ACSP conference, and it was an incredibly enriching experience!

[Thought] Integrating Elements into Geographic Science

10 minute read

Published:

During the past few weeks, I have facilitated two discussion sessions for two elements, one on water and the other on fire. These are two contrasting yet interconnected elements that I have paid close attention to in my study, research, and life over the past years. At this moment, I find myself thinking deeply about how to organize my future studies so that they can better reflect the rudimentary influences of these elements in exploring human–nature interactions through the lens of GeoAI. AI is the hot topic these days that people are flooding into, yet most people end up without really knowing what they are doing with it. I know that I have to start from the elements.