Accurate and efficient AI-assisted paradigm for adding granularity to ERA5 precipitation reanalysis

Abstract

Abstract Scientific inquiry has long relied on deterministic algorithms for systematic problem-solving and predictability. However, the rise of artificial intelligence (AI) has revolutionized data analysis, allowing us to uncover complex patterns in large datasets. In this study, we combine these two approaches by using AI to improve the reconstruction of past precipitation events, which is crucial for understanding climate change. Our objective is to leverage AI to map large-scale atmospheric proxies from the ERA5 climate reanalysis and multi-satellite historical precipitation data from the NASA-IMERG GPM constellation to observed precipitation, enhancing the accuracy and the resolution of climate reanalysis. Accurate climate reanalyses are essential, as they provide the most realistic representations of past atmospheric conditions, serving as benchmarks against which climate models are validated. Our AI-enhanced method offers a more accurate and computationally efficient solution compared to deterministic high-resolution precipitation downscaling methods. Additionally, it shows the capability to generalize predictions to new, previously unobserved locations, making it applicable across various regions. By integrating AI with traditional reanalysis techniques, we open up new opportunities for climate science and geosciences, with the potential to improve the accuracy and reliability of climate data, contributing to a better understanding of climate dynamics.

Keywords

• Past climate reconstruction
• AI and computing in synergy
• Hindcasting