AI天气模型并非普遍在极端天气上表现差，部分模型甚至更优。
Do AI weather models miss extremes?
- 对比11种物理与AI模型，评估其在欧洲多地的极端天气预测能力。
- 多个AI模型在强风、高温、晴天太阳辐射等极端条件下领先，最高提升24.8%。
- 模型缺陷是特定的，非AI整体问题，数值预报模型也有类似偏差。
首批人工智能天气模型常被指在极端天气下表现不佳，主要源于基于再分析数据的确定性回归系统评估。本文将十一类物理与AI预报系统，在欧洲气象站（包括地表、太阳辐射和雨量站）连续十个月的数据上进行验证，评估10米风速、2米气温、小时短波辐射累积及小时降水，以均方误差（MAE）对比ECMWF IFS在ERA5 1991–2020气候区间的表现。结果显示，AI模型在极端尾部并无统一相对技能劣势：Jua EPT-2.1 Europa在全条件风速上领先8.4%，Jua EPT-2 HRRR在温度整体及高温条件下分别领先12.1%和19.6%±2.2%；两者及DWD ICON Global在烈风条件下最优。Jua EPT-2.1 Helios在太阳辐射总体及云层遮蔽和晴空尾部分别领先10.2%±1.7%、16.4%±3.4%和24.8%±5.4%。降水方面，三个Jua模型在中等强度时提升14–15%，在P75–P95区间提升9–11%；而EPT-2 Reasoning在超95百分位仍领先1.7%±0.5%。失败情况为模型特异：ECMWF AIFS在高温尾部下降4.9%±2.0%，NOAA GFS下降22.8%±2.0%。所有模型（含数值预报）均表现出向观测分布中心的共性偏差，模型间差异远小于共同信号。因此，极端天气相对性能缺失并非AI模型共性，而是特定模型问题。
原文摘要 · Abstract (English)
First-generation AI weather models are often reported to underperform at extremes, mostly in reanalysis-based evaluations of deterministic regression systems. We verify eleven physical and AI forecast systems against European synoptic, solar, and rain-gauge stations over ten months for 10 m wind, 2 m temperature, hourly shortwave accumulation, and hourly precipitation, scoring mean absolute error (MAE) against ECMWF IFS in ERA5 1991-2020 climatological regimes. Among these systems, AI models do not show a uniform relative-skill deficit in the tails. Jua EPT-2.1 Europa leads all-conditions wind (+8.4%), while Jua EPT-2 HRRR leads temperature overall (+12.1%) and in the heat regime (+19.6 +/- 2.2%). EPT-2.1 Europa and DWD ICON Global lead at gale-force wind. Jua EPT-2.1 Helios leads solar overall (+10.2 +/- 1.7%), in overcast conditions (+16.4 +/- 3.4%), and in the clear-sky tail (+24.8 +/- 5.4%). For precipitation, three Jua models gain 14-15% at moderate intensity and 9-11% at P75-P95; EPT-2 Reasoning remains ahead above P95 (+1.7 +/- 0.5%). Failures are model-specific: ECMWF AIFS loses 4.9 +/- 2.0% in the heat tail, while NOAA GFS loses 22.8 +/- 2.0% there. Every model, including numerical weather prediction systems, shows a shared conditional bias toward the centre of the observed distribution, with an inter-model spread several times smaller than the shared signal. Missing relative skill at extremes is therefore not a property of AI weather models as a class, but of particular AI and physical models.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。