Supporting material for "Enhancing debris flow warning via machine learning feature reduction and model selection"

Major Changes for v1.2 1, Added v1.2 model based on Type E features (13 selected features) with LSTM + Attention; <br>it may not outperform the Feature C model (2020 Illgraben data) but shows more robust transferability. <br>2, Added test catchments from 2023–2024 Wandong (in-house data):[2023 results](demo/Wandong/pro_Luding-WD-2023-STA02-BHZ-E-testing-False.png), [2024 results](demo/Wandong/pro_Luding-AM-2024-R9BF5-EHZ-E-testing-False.png) <br>3, Added real-time testing for Yanmen catchment (ongoing): [Yanmen](demo/Yanmen)v1.2版本主要更新内容： 1. 新增基于E型特征（13项筛选特征）构建的v2模型，采用长短期记忆网络（LSTM）+注意力（Attention）架构；该模型虽未必优于基于2020年伊尔格拉本(Illgraben)数据集训练的C特征模型，但具备更优异的迁移鲁棒性。 2. 新增2023–2024年旺东（Wandong）流域的测试集数据（内部自有数据集）：[2023年测试结果](demo/Wandong/pro_Luding-WD-2023-STA02-BHZ-E-testing-False.png)、[2024年测试结果](demo/Wandong/pro_Luding-AM-2024-R9BF5-EHZ-E-testing-False.png) 3. 新增雁门（Yanmen）流域的实时测试任务（当前仍在进行中）：[雁门流域测试入口](demo/Yanmen)