Supervised classification of agricultural soil types in Togo using SoilGrids, Google Earth Engine, Random Forest and Leaflet interactive mapping
# Togo Soil Classification
> Supervised classification of agricultural soil types across the 599 administrative cantons of Togo,
> combining SoilGrids v2.0 (ISRIC/FAO), Google Earth Engine Sentinel-2 imagery,
> Random Forest machine learning, and an interactive Leaflet web map.
**Author:** Daniel ESSONANI
**Supervisor:** M. Leri TCHANTCHO
**Date:** February 2026
---
## 🗺️ Live Map
**View the interactive map →**
Each of the 599 cantons is clickable and shows:
- Dominant WRB soil type
- Fertility level
- Recommended crops
- Main agricultural constraints
---
## Project Overview
Togo's agricultural sector employs ~65% of the active population but suffers from a
critical lack of up-to-date soil spatial data — existing pedological maps date from the
1970s–80s and cover only part of the territory.
This project addresses that gap by building a **national-scale soil classification pipeline**
using freely available satellite data and open-source tools, producing results directly
usable by agricultural planners and decision-makers.
The work also serves as the **complementary validation component** of a larger academic
study on ERDAS IMAGINE supervised classification of soil types in Togo
(*"La classification supervisée des types de sols agricoles au Togo grâce à ERDAS IMAGINE"*,
Daniel ESSONANI, NONCHALENT INC., February 2026).
---
## Technical Stack
| Tool | Role |
|------|------|
| Python 3 + Jupyter | Data processing pipeline |
| SoilGrids v2.0 REST API (ISRIC/FAO) | WRB soil type reference labels |
| Google Earth Engine Python API | Sentinel-2 spectral feature extraction |
| scikit-learn | Random Forest classifier |
| GeoPandas | Shapefile processing and spatial joins |
| QGIS + qgis2web | Cartographic styling and Leaflet export |
| Leaflet.js | Interactive web map |
---
## Repository Structure
```
togo-soil-classification/
│
├── README.md
├── rapport_complet_togo_sols.docx ← Full academic report (87 pages, French)
│
├── map/ …