Rapid assimilation and analysis of a suit of remote sensing data for predicting extreme events and their impact on ecological-human systems

Abstract

Focal Areas. Insight gleaned from complex data using AI, and other advanced methods, including explainable AI and physics-or knowledge-guided AI. Predictive modeling through the use of AI techniques and AI-derived model components; the use of AI and other tools to design a prediction system comprising of a hierarchy of models.Science Challenge. Ongoing and future remote sensing (RS) missions are expected to provide an unprecedented amount of data. The hope is that these data will improve our predictive understanding of the Earth System functioning and its response to extreme climate events. This 'data flood', however, poses several challenges: (1) how can we quickly synthesize the volume of data and identify emergent behavior of the ecosystems, particularly under extreme events; and (2) how can this knowledge be effectively incorporated in Earth system models (ESMs) and advance both applied and theoretical research? We envision the next-generation of ESMs will be deeply integrated with theory-informed RSsystems. However, we argue, that the most rapid pathway to get there is to develop a heuristic method based on theory/model-informed RS-based estimation will more rapidly advance scientific discoveries via AI capabilities. Such an approach will be able to identify how to more rapidly assimilate and analyze RS data. Developing such capabilities will be critical in capturing the key drivers of biosphere and ecosystem change -extreme events (e.g., storms, droughts) and associated hazards (e.g., fires, landslides, floods) -and quantifying their impact on both natural (biodiversity, environment) and human (infrastructure, energy, agriculture) systems.