发表机构
Hankuk University of Foreign Studies; Jeju National University; NAVI Hyper-Tropicalization Research Institute（韩国外国语大学; 济州国立大学; NAVI超热带化研究所）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
CLARA模型仅从站点观测学习五个地面变量的联合高斯概率预报，无需数值天气预报，在96个站点上能量分数比基线低4.9%-65%，并在60个区域-前导比较中优于持续性预报57次。
AI 中文摘要
评估复合天气风险需要能够表示变量之间依赖关系的预报。CLARA（校准平流-路由注意力）仅从站点观测学习五个地面变量的联合高斯预测分布，无需数值天气预报或再分析数据；该模型约含28,000个参数，支持在CPU上进行训练和预测。在96个站点的六个多年交叉验证折中，其前导平均能量分数比具有匹配时间输入的已学习比较器低4.9%（参数数量相似时为4.7%），比统计基线低11-65%。在保持边际方差固定的情况下，移除学习到的相关性会使联合负对数似然在每个站点恶化1.0-2.8纳特。一个在所述假设下被证明一致的协方差尺度估计器改善了短前导校准，但在长前导时过度校正。合成干预表明，仅注意力偏置系数不能衡量预报影响。在六大洲的十个区域重新训练后，CLARA在所有60个多年区域-前导比较中优于持续性预报，并在60个中的57个中优于类似规模的已学习模型。
英文摘要
Assessing compound weather risks requires forecasts representing dependence between variables. CLARA (Calibrated Advection-Routing Attention) learns joint Gaussian predictive distributions of five surface variables from station observations alone, without numerical weather prediction or reanalysis; the approximately 28,000-parameter model supports CPU training and prediction. Across six multi-year folds on 96 stations, its lead-mean energy score is 4.9% lower than that of a learned comparator with matched temporal inputs (4.7% with a similar parameter count) and 11-65% lower than those of statistical baselines. Holding marginal variances fixed, removing learned correlations worsens joint negative log-likelihood by 1.0-2.8 nats per station. A covariance-scale estimator, proved consistent under stated assumptions, improves short-lead calibration but over-corrects at long leads. Synthetic interventions show an attention-bias coefficient alone does not measure forecast influence. Retrained in ten regions on six continents, CLARA outperforms persistence in all 60 multi-year region-lead comparisons and a similarly sized learned model in 57 of 60.
Comments62 pages, 6 figures, 3 Extended Data figures, 7 Extended Data tables; includes Supplementary Information
