用随机扰动让冻结的天气模型模拟出稳定气候，100年不漂移。
Rescene: band-limited stochastic forcing turns a frozen neural weather operator into a climate emulator
- 给冻结的神经天气模型加个带频谱限制的随机扰动头，维持长期稳定。
- 恢复了90天内82%的阻塞事件频率，日变率达观测值的1.3倍。
- 适合需要长期气候模拟且不想重训模型的研究者使用。
近年来，机器学习天气预报模型已达到甚至超过欧洲中期天气预报中心（ECMWF）高分辨率预报（HRES）的中程预测能力。然而，这些模型在训练时长之外自由积分时会崩溃、漂移或丢失季节循环，重新训练成本高昂。我们探索能否从一个严格冻结的骨干模型中恢复有用信息。提出Rescene：一个0.4百万参数的封装，基于1.5度、每6小时一次的视觉变换器操作器，利用ERA5再分析数据构建。该模型包含一个0.33百万参数的确定性‘慢钟’，将预报逐步引导至依赖预报时序的日期气候态；以及一个0.06百万参数的生成式头部，每步添加谱形限定的随机扰动。评估显示，仅用确定性部分可稳定运行数十年，但日变率降至ERA5的40%。加入生成头部后，恢复了126%（Z500）和130%（MSLP）的日变率，模式相关系数达0.89和0.92，重现82%的阻塞频率，集合校准良好（第7至第90天的展宽-技能比为0.78–0.97），并可连续积分100年无显著漂移（+0.008 ± 0.014 K/世纪）。值得注意的是，扰动仅作用于总波数 $k \le 20$，小尺度未被强迫，但 $k \ge 20$ 的能量仍保持合理：6小时能量收支分解显示，在 $k \ge 40$ 处，冻结操作器提供的能量是扰动的28倍，网格尺度的相对增长率是行星尺度的247倍。
原文摘要 · Abstract (English)
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.
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。