A Possible Artificial Intelligence Ecosystem Avatar: the Moorea case (IDEA)
This work aims to improve the understanding and prediction of complex island ecosystems for researchers and policymakers by integrating diverse data types into a unified computational framework.
This paper proposes a large-scale data assimilation and prediction backbone for ecosystem understanding, utilizing Deep Stacking Networks (DSN) within the Island Digital Ecosystem Avatars (IDEA) project for Moorea Island. The system integrates multimodal data, including physics, chemistry, biology, ecology, fishing, economics, and social sciences, by subdividing the island into watersheds and lagoon units.
High-throughput data collection techniques and largescale (cloud) computing are transforming our understanding of ecosystems at all scales by allowing the integration of multimodal data such as physics, chemistry, biology, ecology, fishing, economics and other social sciences in a common computational framework. We focus in this paper on a large scale data assimilation and prediction backbone based on Deep Stacking Networks (DSN) in the frame of the IDEA (Island Digital Ecosystem Avatars) project (Moorea Island), based on the subdivision of the island in watersheds and lagoon units. We also describe several kinds of raw data that can train and constrain such an ecosystem avatar model, as well as second level data such as ecological or physical indexes / indicators.