Data from MIKE 21 models for training and validation of Sparse GP models in "Upskilling low-fidelity hydrodynamic models of flood inundation through spatial analysis and Gaussian Process learning"

Data from MIKE 21 models for training and validation of Sparse GP models.The data is used for publication in "Refining l ow-fidelity hydrodynamic model outputs of flood inundation through spatial analysis and Gaussian Process learning" with the Chowilla floodplain as case study.<br><br>The raw data is structured in two folders. One for the high-fidelity model (HF) and one for the low-fidelity model (LF). Both folders contain MIKE 21 .dfsu data files.<br>The managed data is structured in three folders. "Classification_Figures" contain figures generated for the publication. "Events_data" contains the MIKE 21 data in binary format as .npz files to be read via the Numpy package in Python. "SPGP_class_models" contains the trained Sparse GP (SPGP) models, EOF analysis data and categories depending on the binary state of the data on cell level.<br>Python code to generate the managed data is available at https://github.com/nfraehr/Hybrid_LSG_model

本数据集包含用于稀疏高斯过程（Sparse Gaussian Process, SGP）模型训练与验证的MIKE 21模型模拟数据，相关成果已发表于论文《基于空间分析与高斯过程学习优化洪水淹没低保真水动力模型输出结果》（Refining low-fidelity hydrodynamic model outputs of flood inundation through spatial analysis and Gaussian Process learning），研究案例选取乔伊拉泛滥平原（Chowilla floodplain）。<br><br>原始数据分为两个文件夹：其一对应高保真模型（High-Fidelity Model, HF），其二对应低保真模型（Low-Fidelity Model, LF），两个文件夹内均存储MIKE 21格式的.dfsu数据文件。<br>处理后的数据分为三个文件夹："Classification_Figures"文件夹存储用于论文制作的可视化图表；"Events_data"文件夹存储以二进制格式封装的MIKE 21数据，文件格式为.npz，可通过Python的NumPy库读取；"SPGP_class_models"文件夹存储训练完成的稀疏高斯过程（Sparse Gaussian Process, SPGP）模型、经验正交函数（Empirical Orthogonal Function, EOF）分析数据，以及基于单元格级数据二进制状态生成的分类标签。<br>生成该处理后数据的Python代码可在以下链接获取：https://github.com/nfraehr/Hybrid_LSG_model