TN Prediction Results for Miho Stream Watershed using SWAT-DL

# SWAT-ML and Deep Learning Model Results for TN Prediction in Miho Watershed

This dataset contains simulation results and training datasets related to the prediction of Total Nitrogen (TN) concentrations using a hybrid SWAT-ML model and deep learning models (LSTM, GRU) for the Miho watershed.

## 1. Files Included

### [A] SWAT-ML Model Results (Outlet)
- tn_train_pred_0323.csv: Training period predictions for TN at the outlet.
- Columns: Date2, TN
- tn_test_pred_0323.csv: Testing period predictions for TN at the outlet.
- Columns: Date2, TN

### [B] Model Input Dataset
- Miho10_dataset_mod.csv: Input dataset used for training the SWAT-ML model. Includes input features such as SWAT outputs and previous precipitation.
- Columns: Date, SWAT_Flow, SWAT_Sed, SWAT_ORGN, SWAT_ORGP...

### [C] LSTM Model Results (Node-wise)
- Miho3_results_LSTM.csv: LSTM-based TN prediction results at Miho3 site.
- Columns: Unnamed: 0, org, pred
- Miho5_results_LSTM.csv: LSTM-based TN prediction results at Miho5 site.
- Columns: Unnamed: 0, org, pred
- Miho6_results_LSTM.csv: LSTM-based TN prediction results at Miho6 site.
- Columns: Unnamed: 0, org, pred
- Miho9_results_LSTM.csv: LSTM-based TN prediction results at Miho9 site.
- Columns: Unnamed: 0, org, pred

### [D] GRU Model Results (Node-wise)
- Miho3_results_GRU.csv: GRU-based TN prediction results at Miho3 site.
- Columns: Unnamed: 0, org, pred
- Miho5_results_GRU.csv: GRU-based TN prediction results at Miho5 site.
- Columns: Unnamed: 0, org, pred
- Miho6_results_GRU.csv: GRU-based TN prediction results at Miho6 site.
- Columns: Unnamed: 0, org, pred
- Miho9_results_GRU.csv: GRU-based TN prediction results at Miho9 site.
- Columns: Unnamed: 0, org, pred

## 2. Notes
- All prediction results are saved in CSV format, containing at least a Date column and predicted TN values.
- This dataset is part of a study on integrating process-based models (SWAT) with machine learning (ML/DL) models to improve nitrogen load prediction accuracy across multiple locations in a watershed.

## 3. Contact
For questions regarding the dataset, please contact:

**Yongeun Park**
Professor. Department of Civil and Environmental Engineering,
Konkuk University-Seoul, Republic of Korea
Email: [e-mail]