KNOU-Flood: Advancing global streamflow forecasting with lead-aware probabilistic machine learning

Computer Science > Machine Learning arXiv:2610.20640 (cs) [Submitted on 2 Oct 2026] Title: KNOU-Flood: Advancing global streamflow forecasting with lead-aware probabilistic machine learning Authors: Minjong Cheon.

计算机科学 > 机器学习 arXiv:2610.20640 (cs) [提交于 2026 年 10 月 2 日] 标题：KNOU-Flood：利用考虑提前期的概率机器学习推进全球径流预报 作者：Minjong Cheon。

Abstract: Global streamflow forecasting supports flood preparedness, yet data-driven models trained on reanalysis lose skill when driven by operational forecasts, increasingly so with lead time. AIFL addresses this shift with a two-stage transfer-learning recipe and a deterministic LSTM, but it feeds one forecast lead time into the whole input window and gives no uncertainty. We ask how far this open recipe can be extended without changing its recurrent core.

摘要：全球径流预报为防洪准备提供支持，然而，基于再分析数据训练的数据驱动模型在由业务预报驱动时会丧失准确性，且随着提前期的增加，这种损失愈发明显。AIFL 通过两阶段迁移学习方案和确定性 LSTM 来应对这种偏差，但它将单一的预报提前期输入到整个输入窗口中，且不提供不确定性评估。我们探讨了在不改变其循环核心的情况下，这一开放方案能被扩展到何种程度。

KNOU-Flood keeps a single-layer LSTM and the two-stage training and changes the information around it: a longer input window (424 days), lead-aware forecast inputs, GraphCast as a second forecast source next to ECMWF HRES, and a latent-noise ensemble trained with the fair CRPS. We compare against a re-implementation of the AIFL recipe (its code is not public) under identical data, stations, splits and metrics.

KNOU-Flood 保留了单层 LSTM 和两阶段训练过程，并改变了其周围的信息：更长的输入窗口（424 天）、考虑提前期的预报输入、除 ECMWF HRES 外引入 GraphCast 作为第二个预报源，以及使用公平 CRPS 训练的潜在噪声集成模型。我们在相同的数据、站点、划分和指标下，与 AIFL 方案的重新实现版本（其代码未公开）进行了对比。

Our contributions are threefold. First, over three seeds and 2,099 test basins (2021-2023), KNOU-Flood reaches a median KGE’ of 0.65 at LT1 and 0.48 at LT10 against 0.62 and 0.34, with significant gains at all ten lead times under two scoring protocols, also against a 1024-unit baseline and on held-out basins (one seed each).

我们的贡献有三点。首先，在三个随机种子和 2,099 个测试流域（2021-2023 年）的测试中，KNOU-Flood 在 LT1 和 LT10 的中位数 KGE’ 分别达到 0.65 和 0.48，而对比组为 0.62 和 0.34；在两种评分协议下，其在所有十个提前期内均有显著提升，且优于 1024 单元的基准模型及留出流域的测试结果（每个种子各一个）。

Second, identical weights give an LT10 KGE’ of 0.18 or 0.34 depending on how forecasts are fed, so much of the lead-time decay reported for AIFL is an input-protocol effect; the probabilistic head contributes most to the gain, followed by the raw GraphCast channels, the longer window and mixed fine-tuning.

其次，相同的权重在不同的预报输入方式下会产生 0.18 或 0.34 的 LT10 KGE’，这表明 AIFL 报告的提前期衰减很大程度上是输入协议效应所致；概率头对增益的贡献最大，其次是原始 GraphCast 通道、更长的窗口和混合微调。

Third, the ensembles reach lower CRPS than probabilistic versions of the baseline, are nearly calibrated with inference-time dropout, and detect 1.5- and 2-year floods better than all baselines; the benefit for rarer floods depends on the seed, and an exploratory tail-weighted variant improves 10-year and larger events. KNOU-Flood is a controlled extension of an open recipe, not a new architecture.

第三，集成模型比基准模型的概率版本达到了更低的 CRPS，通过推理时 Dropout 实现了近乎校准，并且在检测 1.5 年和 2 年一遇洪水方面优于所有基准模型；对更罕见洪水的增益取决于随机种子，而一种探索性的尾部加权变体改善了 10 年及以上量级事件的预报。KNOU-Flood 是对开放方案的受控扩展，而非一种全新的架构。