Uncertainty-Aware End-to-End AI Weather Forecasting: Disentangling Observation and Model Contributions

• 类型：arxiv
• 标识：2608.30795
• 链接：https://arxiv.org/abs/2608.30795
• 主分类：llm-infra
• 形态：method
• 被引：0
• 被引来源：Semantic Scholar + OpenAlex
• S2被引：0
• OpenAlex被引：0
• 影响力被引：0
• TLDR：End-to-end weather forecasting systems produce skillful global gridded and station forecasts directly from raw Earth observations, replacing the numerical weather prediction pipeline, including data assimilation, at a fraction of its cost. These systems are deterministic and issue no uncertainty. Here we render the Aardvark Weather model probabilistic by attaching one stochastic mechanism to each component: learned, input-dependent noise at the observation encoder, capturing aleatoric uncertainty inherited from the observing system, and Monte Carlo dropout in the processor, capturing epistemic
• OpenAlex ID：W7204966142
• OpenAlex DOI：10.48550/arxiv.2608.30795
• DOI：10.48550/arxiv.2608.30795
• DOI来源：OpenAlex
• 开放获取：green
• 开放获取链接：https://arxiv.org/pdf/2608.30795
• OpenAlex更新：2026-09-04
• 待LLM分类：否
• 标题中文：感知不确定性的端到端 AI 天气预报：分离观测与模型的贡献
• TLDR中文：端到端天气预报系统直接从原始地球观测数据生成高质量的全球网格与站点预报，仅以数值天气预报流水线（含数据同化）一小部分的成本取而代之。这类系统是确定性的，不输出不确定性结果。本文通过在每个组件上附加一种随机机制，将 Aardvark Weather 模型概率化：在观测编码器中加入学习到的、依赖输入的噪声，以捕捉源自观测系统的偶然不确定性；在处理器中使用 Monte Carlo dropout，以捕捉认知不确定性。
• 来源文件：
• /inbox/tom/_candidates/2026-09-02-agent-rag-longcontext-candidates.json
• [S2 enrich]
• [OpenAlex backfill]