A new framework for Observing System Simulation Experiments (OSSEs) has been developed to evaluate the physical consistency of AI weather models. This framework was used to test four distinct AI models: FourCastNet V2, GraphCast, Aurora, and Prithvi WxC, over a 60-day forecast horizon. The study found that different AI architectures respond variably to observational data, with foundation models like Prithvi WxC and Aurora showing the most significant impact. However, an architecture-specific numerical instability was identified in GraphCast's full configuration, which could be mitigated by adjusting the model's vertical levels. AI
IMPACT This research introduces a new method for evaluating AI weather models, highlighting architectural differences and potential instabilities, which could inform future model development.
RANK_REASON The cluster contains an academic paper detailing a new framework and evaluation of AI weather models. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →