Advancing Machine Learning Tools for Earth Science: Workshop Report

NASA and Radiant Earth Foundation recently held a workshop that brought together machine learning practitioners and domain experts to discuss how to effectively use machine learning techniques with Earth Observation data. The report from this workshop is now available.
author-share

In January 2020, NASA Earth Science Data Systems (ESDS) and the Radiant Earth Foundation hosted a workshop for experts to discuss the advancement of machine learning (ML) techniques on NASA’s Earth Observation (EO) data. Held in Washington D.C., the workshop included 51 participants from government agencies, non-profit organizations, universities, and private industries. The report from this workshop is now available (PDF), which highlights the challenges, potential solutions, and best practices for using EO data in ML workflows for Earth science research and applications.

 

Image
Participants of the machine learning workshop are shown sitting around tables.
The Advancing Machine Learning Tools for Earth Science Workshop included 51 participants from government agencies, non-profit organizations, universities, and private industries. Image courtesy of Radiant Earth Foundation.

Machine learning is the branch of artificial intelligence that can learn from data, identify patterns, and make decisions with minimal human intervention. Earth science disciplines are especially primed to take advantage of ML because of the wealth of readily available EO data.

The building blocks for widespread adoption of ML in Earth science include open data, open-source software, community building, research in advanced algorithm development, and benchmark-labeled datasets. Towards that end, NASA ESDS program has invested in ML-based research and applications focusing on data-driven science and improving operational efficiencies. Efforts are also underway to generate highly curated benchmark training Earth science datasets that can be used to accelerate advanced algorithm development and benchmarking. These activities, along with NASA’s open data and open source policies, place ESDS in a strong position to take full advantage of opportunities that ML presents. As ESDS transitions its data and operations to the commercial cloud, the readily available cloud services next to the data provide opportunities to scale ML applications.

But challenges remain in adopting ML for Earth science, including lack of training datasets and transitioning ML applications from research to production. Workshop participants discussed these key challenges and provided recommendations on how to effectively move forward.

A list of presenters and participants is available on the workshop website, and recorded presentations are available on YouTube. Read the report from the expert workshop: Advancing Application of Machine Learning Tools for NASA’s Earth Observation Data (PDF).

More Information

Learn more about NASA’s Open Data, Services, and Software Policies.

Last Updated