Nile River Sudan MultiSite ML Dataset

This repository contains the **Nile Sudan MultiSite ML Dataset**, a machine-learning-ready geospatial dataset generated using Google Earth Engine (GEE) for hydrological and remote sensing analysis in Sudan along the Nile River system.

The dataset was developed for supervised machine learning applications related to:
- flood mapping,
- water-body detection,
- hydrological monitoring,
- and environmental classification.

---

## Study Area

The dataset focuses on selected Nile River regions in Sudan, including areas surrounding the Merowe region and associated hydrological zones.

Primary study region:
- Merowe Dam region, Sudan

---

## Data Sources

The dataset was generated using multiple remote sensing and terrain products available in Google Earth Engine:

### Satellite Data
- Sentinel-1 SAR imagery
- Sentinel-2 multispectral imagery

### Terrain and Hydrological Data
- Digital Elevation Model (DEM)
- Terrain derivatives:
  - elevation
  - slope
  - aspect
  - flow accumulation

### Climate Data
- Rainfall products extracted from GEE climate datasets

---

## Dataset Features

The dataset includes the following predictor variables:

| Feature | Description |
|---|---|
| NDVI | Normalized Difference Vegetation Index |
| NDWI | Normalized Difference Water Index |
| MNDWI | Modified NDWI |
| VV | Sentinel-1 SAR backscatter |
| blue | Sentinel-2 blue band |
| green | Sentinel-2 green band |
| red | Sentinel-2 red band |
| nir | Near Infrared band |
| swir1 | Shortwave Infrared 1 |
| swir2 | Shortwave Infrared 2 |
| elevation | Terrain elevation |
| slope | Terrain slope |
| aspect | Terrain aspect |
| flow_acc | Flow accumulation |
| rainfall_mean | Mean rainfall |
| rainfall_sum | Total rainfall |
| label | Classification label |
| random | Randomized split variable |

---

## Coordinate System

The dataset was generated within the Google Earth Engine environment using standard geographic coordinate systems associated with Sentinel products.

Coordinate information is stored in:
- `.geo`

---

## Temporal Coverage

The dataset was generated from multi-temporal satellite observations processed within Google Earth Engine.

---

## Preprocessing Workflow

The preprocessing pipeline included:
1. Satellite image acquisition
2. Cloud filtering and preprocessing
3. Spectral index calculation
4. Terrain feature extraction
5. Rainfall variable integration
6. Sample extraction
7. Randomized train/test preparation
8. CSV export from Google Earth Engine

---

## Machine Learning Applications

This dataset is suitable for:
- Random Forest classification
- XGBoost models
- Support Vector Machines (SVM)
- Deep learning workflows
- Flood susceptibility mapping
- Water classification
- Environmental monitoring

---

## Software

Dataset generation was performed using:
- Google Earth Engine (GEE)