Predicting submesoscale tracer dispersion with neural networks

This record contains data used in the paper "Predicting submesoscale tracer dispersion with neural networks". It contains: Vort_Trac folder: 1) Snapshots of bartropic vorticity of many oceanic flow simulations across varying Rossby number and energy in baroclinic instability 2) Snapshots of tracer fields advected in the same flows. FigData_numpy folder: 3) Model predictions used in the figures and analysis in the paper. Files are named as: z_{Rossby_number}_{baroclinic_energy}_{time}.npy for vorticity snapshots tr_{Rossby_number}_{baroclinic_energy}_{time}.npy for tracer snapshots Other files are name in a similar way. 'ModelAndDerivations.ipynb' is a jupyter notebook file that defines all the models and has functions to load parameters from the 'models' folder. It also contains functions to derive spectra, flux, spectral loss, etc 'models' folder contained parameters of the trained neural networks Vtrsimulation.py is python file that was used to preduce Flow and tracer fields numerically.本数据集包含论文《基于神经网络预测亚中尺度示踪剂弥散》（"Predicting submesoscale tracer dispersion with neural networks"）所使用的实验数据，具体内容如下： ### Vort_Trac 文件夹 1. 多组海洋流场模拟的正压涡度（bartropic vorticity）快照，模拟场景覆盖斜压不稳定（baroclinic instability）下不同罗斯比数（Rossby number）与能量参数的组合 2. 相同流场中经平流输运的示踪场（tracer fields）快照 ### FigData_numpy 文件夹 3. 论文绘图与分析环节所使用的模型预测结果 #### 文件命名规则 - 涡度快照文件命名格式为 `z_{Rossby_number}_{baroclinic_energy}_{time}.npy` - 示踪场快照文件命名格式为 `tr_{Rossby_number}_{baroclinic_energy}_{time}.npy` - 其余文件命名规则与此类似 ### 配套文件说明 - `ModelAndDerivations.ipynb` 为 Jupyter Notebook 文件，定义了全部模型，并提供了从 `models` 文件夹加载已训练神经网络（neural networks）参数的函数；同时内置了计算光谱（spectra）、通量（flux）、谱损失（spectral loss）等的相关功能函数 - `models` 文件夹存储了已训练神经网络的参数文件 - `Vtrsimulation.py` 为用于数值生成流场与示踪场的 Python 脚本文件