Community-Driven Development of Tools to Improve AI-Readiness of the Open Environmental Data

As artificial intelligence (AI) and machine learning (ML) gaining broad interests in the Earth and space science community, the demand for AI-ready data can support the development of responsible AI/ML applications with open environmental data. Through a broad community collaboration under Earth Science Information Partners, we have developed an AI-readiness checklist as a community guideline for the development of AI-ready open environmental data. The checklist was initially based on an early draft of an AI-ready matrix developed by the OSTP Open Science Sub-committee but has been modified notably based on feedback from data users and AI/ML practitioners. The current version of the AI-readiness checklist can be used to holistically assess the documentation, quality, access, and pre-processing of a given dataset. The AI-readiness assessment result can be then summarized into a data card that provides human-readable metrics to assist users in determining if the dataset meets the user's need for their AI/ML development. The next milestone of this community-driven effort is to develop a community-driven convention by building on the existing data conventions and standards to fill the data management gap to support AI-ready data management. In this presentation, we will also showcase a collection of AI-ready climate datasets applying the AI-readiness checklist and data card concept to support AI/ML applications in climate sciences. The AI-readiness development process requires active community engagement with data repositories, domain scientists, and AI/ML practitioners to establish a flexible framework to ensure the rapid evolution of AI/ML technologies can be addressed in modern data management.