Submarine Landslide Risk Concerning Military Conflicts in the Strait of Hormuz and Gulf of Oman

Environmental Features This folder contains the raster files of the six environmental variables used as input features for the machine learning models: Water Depth (`Depth.tif`), Slope (`Slope.tif`), Roughness (`Roughness.tif`), Curvature (`Curvature.tif`), Distance to Fault (`Fault.tif`), and Peak Ground Acceleration (`PGA.tif`). The `Depth.xyz` file contains the raw coordinate and depth data. All raster files are provided in GeoTIFF format with a spatial resolution of approximately 450 m (15 arc seconds), covering the study area in the Strait of Hormuz and the northern Gulf of Oman. Model Weights This folder contains the trained model weights for the four ensemble learning algorithms evaluated in this study: `RandomForest_model.pkl`, `XGBoost_model.pkl`, `LightGBM_model.pkl`, and `CatBoost_model.pkl`. The models were trained using the environmental features listed above and the global submarine landslide inventory. The weights are saved in Python pickle (.pkl) format and can be loaded using the corresponding libraries (scikit-learn, XGBoost, LightGBM, CatBoost). Prediction Results This folder contains the submarine landslide susceptibility prediction results generated by the optimal Random Forest model under nine scenarios: Background (no explosion), the Strait of Hormuz (10 kt), and the Gulf of Oman (1 t, 10 t, 100 t, 1 kt, 10 kt, 100 kt, and 1 Mt). Each raster file (`Prediction_*.tif`) contains the predicted landslide susceptibility probability (ranging from 0 to 1) at a spatial resolution of approximately 450 m in GeoTIFF format. Earthquake Location This folder contains 8 text files (`Gulf of Oman PGA_*.txt` and `Strait of Hormuz PGA_100kt.txt`) specifying the location (longitude, latitude), magnitude, and focal depth of simulated explosions. Each file contains a single line of data used as input for PGA calculation and subsequent landslide susceptibility modeling. Dataset This folder contains the training dataset (`Training Dataset.txt`) and prediction datasets (`predicted dataset_*.txt`) used for model development and scenario simulations. All files are tab-delimited text files with consistent feature columns, serving as input for landslide susceptibility prediction. Python Scripts The following Python scripts are located in the root directory and were used for data processing, model training, and prediction: - `Nuclear_explosion_seismology.py` – Calculates the equivalent earthquake magnitude for explosions at different locations and yields. - `Machine_learning_classification_model_training_and_prediction.py` – Performs model training and generates landslide susceptibility predictions. - `Generate_Prediction_Dataset.py` – Prepares the input datasets required for prediction. - `Calculate_PGA.py` – Computes the Peak Ground Acceleration (PGA) across the entire study area following simulated explosions.环境特征 本文件夹包含用于机器学习模型 (machine learning models) 输入特征的6种环境变量的栅格文件 (raster files)：水深（`Depth.tif`）、坡度（`Slope.tif`）、粗糙度（`Roughness.tif`）、曲率（`Curvature.tif`）、距断层距离（`Fault.tif`）以及地面峰值加速度 (Peak Ground Acceleration，PGA)（`PGA.tif`）。`Depth.xyz` 文件包含原始坐标与水深数据。所有栅格文件均采用GeoTIFF格式，空间分辨率约为450米（15角秒），覆盖霍尔木兹海峡与阿曼湾北部的研究区域。 模型权重 本文件夹包含本研究评估的4种集成学习算法 (ensemble learning algorithms) 的训练后模型权重：`RandomForest_model.pkl`、`XGBoost_model.pkl`、`LightGBM_model.pkl` 以及 `CatBoost_model.pkl`。上述模型均基于前文列出的环境特征与全球海底滑坡编目 (submarine landslide inventory) 进行训练。权重以Python pickle（.pkl）格式存储，可通过对应库（scikit-learn、XGBoost、LightGBM、CatBoost）加载。 预测结果 本文件夹包含最优随机森林模型在9种情景下生成的海底滑坡敏感性预测结果：背景情景（无爆炸）、霍尔木兹海峡（10千吨当量，kt）、阿曼湾（1 t、10 t、100 t、1 kt、10 kt、100 kt 以及1 Mt当量）。每个栅格文件（`Prediction_*.tif`）均包含预测的滑坡敏感性概率（取值范围为0至1），空间分辨率约为450米，采用GeoTIFF格式。 地震位置 本文件夹包含8个文本文件（`Gulf of Oman PGA_*.txt` 与 `Strait of Hormuz PGA_100kt.txt`），用于指定模拟爆炸的位置（经度、纬度）、震级与震源深度。每个文件仅包含一行数据，用于PGA计算与后续滑坡敏感性建模。 数据集 本文件夹包含用于模型开发与情景模拟的训练数据集（`Training Dataset.txt`）与预测数据集（`predicted dataset_*.txt`）。所有文件均为制表符分隔的文本文件，特征列保持一致，可作为滑坡敏感性预测的输入数据。 Python脚本 根目录下存放了以下用于数据处理、模型训练与预测的Python脚本： - `Nuclear_explosion_seismology.py`：计算不同位置与当量的核爆炸的等效地震震级。 - `Machine_learning_classification_model_training_and_prediction.py`：执行模型训练并生成滑坡敏感性预测结果。 - `Generate_Prediction_Dataset.py`：生成预测所需的输入数据集。 - `Calculate_PGA.py`：计算模拟爆炸后整个研究区的地面峰值加速度（PGA）。