I use machine learning and remote sensing to measure greenhouse gas emissions and land use change, both today and in the past. Below is an overview of my main research projects. Papers are listed on the Publications page.

Methane quantification from space

University of Washington, 2026 – present

As a Postdoctoral Fellow in the Turner Group, I work on advancing methane quantification using remote sensing data. This work is supported by the Gordon and Betty Moore Foundation.

Measuring CO₂ emissions from power plants with deep learning

Global Policy Lab, 2021 – 2026

Power plants are among the largest point sources of carbon dioxide, but direct measurements of their emissions are sparse. We explore deep learning methods that estimate CO₂ emissions from individual power plants using satellite observations, working toward a framework that generalizes across plants and regions.

  • Murayama, H., Wang, S., Sherman, L., Cohen, R., Hsiang, S. Parameterizing deep learning for carbon dioxide emission quantification for power plants. In preparation.

Reconstructing historical land cover in Africa

Global Policy Lab, 2021 – 2026

Our picture of how African landscapes have changed over the past century is limited by the lack of historical data. Using an archive of roughly one million historical aerial photographs, I]we use deep learning to build high-resolution, historical land cover maps across Africa.

  • Murayama, H., Noda, E., Chong, T., et al. Developing high resolution, historical land cover maps in Africa using deep learning and an aerial photography archive. In preparation.

Machine learning with satellite imagery

I contributed to a global evaluation of how well machine learning with satellite imagery can measure ground conditions across a wide range of variables.

Earlier work