Data and Code for 'Integrating Physics-Informed and Data-Driven Neural Networks into Earth System Models: A Comparative Study for Compound Flood Simulation at River-Ocean Interfaces'

Installation and Requirements<br><br>PINNs (Physics-Informed Neural Networks)<br>TensorFlow Version: Requires TensorFlow 1.14.0<br>Environment Setup: Use the environment details provided in `requirement_tf1.txt` to set up your Conda environment.<br>conda create --name tf1 --file requirement_tf1.txt<br>conda activate tf1<br><br><br>Data-driven Model<br>TensorFlow Version: Requires TensorFlow 2.17.0<br>Environment Setup: Use the environment details provided in `requirement_tf2.txt` to set up your Conda environment.<br><br>conda create --name tf2 --file requirement_tf2.txt<br>conda activate tf2<br><br><br>### Before training<br>Before running the code, need to create folders to save the model output<br><br>For CNN, create /files/CNN<br><br>For PINNs, create /saved_model<br><br>For saving figures from visualization, create /figures<br><br>Training and Results<br>PINNs<br>Training: To train the model, run:python PINN_test_bnd_uh_Telemac.pypython PINN_test_bnd_uh_Telemac_FDM.py<br>Result Plotting and Comparison: For plotting and comparing results, use:python PINN_plot_comparison.py<br><br><br>Data-driven Model<br>CNN Training: To train the CNN model, execute:python train_CNN.py<br><br>Result Visualization: To visualize the results of the CNN model, run:python predict_CNN.py<br><br>To reproduce all results and figures in the manuscript, please refer to the scripts in analysis/

安装与环境要求

物理信息神经网络（Physics-Informed Neural Networks，简称PINNs）
TensorFlow版本要求：需使用TensorFlow 1.14.0
环境配置：通过`requirement_tf1.txt`中提供的依赖信息配置Conda环境，具体命令如下：
conda create --name tf1 --file requirement_tf1.txt
conda activate tf1

数据驱动模型
TensorFlow版本要求：需使用TensorFlow 2.17.0
环境配置：通过`requirement_tf2.txt`中提供的依赖信息配置Conda环境，具体命令如下：
conda create --name tf2 --file requirement_tf2.txt
conda activate tf2

### 训练前准备
运行代码前，请创建用于存储模型输出结果的文件夹：
- 对于卷积神经网络（Convolutional Neural Network，简称CNN），需创建`/files/CNN`目录
- 对于PINNs，需创建`/saved_model`目录
- 如需保存可视化生成的图像，请创建`/figures`目录

训练与结果生成
PINNs
训练：执行以下命令以训练模型：
`python PINN_test_bnd_uh_Telemac.py`、`python PINN_test_bnd_uh_Telemac_FDM.py`
结果绘图与对比：如需绘制结果并进行对比分析，请执行：
`python PINN_plot_comparison.py`

数据驱动模型
CNN训练：如需训练CNN模型，请执行：
`python train_CNN.py`
结果可视化：如需可视化CNN模型的输出结果，请执行：
`python predict_CNN.py`

如需复现论文手稿中的全部结果与图像，请参考`analysis/`目录下的脚本文件。