发表机构
Sejong University（世宗大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究提出Rescene包装器，通过带限随机强迫使冻结的神经天气算子稳定运行百年，恢复了观测的日变异性、阻塞频率等关键气候特征，为低成本构建气候模拟器提供了方案。
AI 中文摘要
过去几年，用于天气预报的机器学习（ML）模型发展迅速，产生的确定性模型的中期预报技巧已达到或超过欧洲中期天气预报中心（ECMWF）的高分辨率预报（HRES）。然而，当这些模型在其训练的预报时效之外自由积分时，会出现崩溃、漂移或失去季节循环的问题，而重新训练它们以获得稳定性的成本很高。因此，我们研究能否从严格冻结的骨干模型中恢复所需性能。我们提出了Rescene，这是一个参数规模为0.4M的包装器，围绕一个冻结的1.5度、6小时分辨率的视觉Transformer算子构建，使用ERA5再分析数据开发，包含一个确定性的“慢时钟”（0.33M参数），用于将预报向基于预报时效的年际气候态融合，以及一个生成式头部（0.06M参数），在每一步添加频谱整形的随机扰动。性能评估表明，仅确定性包装器本身可稳定运行数十年，但会将日变异性降低至ERA5的40%；添加生成式头部后，可恢复观测到的日变异性的126%（Z500）和130%（MSLP），模式相关系数分别为0.89和0.92，恢复了观测到的阻塞频率的82%，使集合保持校准状态（从第7天到第90天的 spread-skill 比率为0.78-0.97），并可积分100年且无明显漂移（每世纪+0.008±0.014 K）。此外，由于扰动被带限至总波数k≤20，小尺度从未被强迫，但维持了现实的k≥20功率：对6小时能量收支的直接分解显示，冻结算子在k≥40时提供的能量比扰动多28倍，在网格尺度的分数增长率比行星尺度大247倍。
英文摘要
Over the past few years, the rapid development of machine learning (ML) models for weather forecasting has produced deterministic models whose medium-range skill matches or exceeds that of the European Centre for Medium-Range Weather Forecasts (ECMWF)'s high-resolution forecast (HRES). However, when these models are integrated freely beyond the horizon they were trained for, they blow up, drift, or lose their seasonal cycle, and retraining them for stability is expensive. We therefore ask what can be recovered from a strictly frozen backbone. We present Rescene, a 0.4 M-parameter wrapper around a frozen 1.5 degree, 6-hourly vision-transformer operator, developed using ERA5 reanalysis data and comprising a deterministic "slow clock" (0.33 M) that blends the forecast toward a lead-aware day-of-year climatology and a generative head (0.06 M) that adds a spectrally shaped stochastic perturbation at every step. The performance evaluation demonstrates that the deterministic wrapper alone is stable for decades but collapses daily variability to 40% of ERA5. Adding the generative head restores 126% (Z500) and 130% (MSLP) of the observed daily variability with pattern correlations of 0.89 and 0.92, recovers 82% of the observed blocking frequency, keeps the ensemble calibrated (spread-skill ratio 0.78-0.97 from day 7 to day 90), and integrates for 100 years with no detectable drift (+0.008 +/- 0.014 K per century). Moreover, because the perturbation is band-limited to total wavenumber $k \le 20$, the small scales are never forced, yet realistic $k \ge 20$ power is sustained: a direct decomposition of the 6-hourly energy budget shows that the frozen operator supplies 28 times more energy than the perturbation at $k \ge 40$, with a fractional growth rate 247 times larger at the grid scale than at planetary scales.
