gayathry32/Cocoa-Yield-Prediction-in-Ghana-using-Remote-Sensing-Random-Forest-2000-2023
Domain:
agricultureRecord type:
project
Creator:
gay
Host:
Cocoa yield prediction for Ghana (2000–2023) using FAOSTAT data with CHIRPS rainfall, ERA5 temperature, and MODIS NDVI. Features aligned to cocoa phenology and modeled via Random Forest with time-based validation (train: 2000–2018, test: 2019–2023). Achieved ~62% R², supporting early warning and trading signal applications.
# 🌍 Cocoa Yield Prediction – Ghana (2000–2023)
## 📌 Project Overview
This project develops a **cocoa yield prediction model for Ghana (2000–2023)** by integrating
**FAOSTAT yield data** with seasonal **satellite and climate indicators**:
- 🌧️ **CHIRPS Rainfall** (May–Aug, flowering & pod set)
- 🌡️ **ERA5-Land Temperature** (May–Aug, heat stress)
- 🌱 **MODIS NDVI** (Aug–Oct, pod filling canopy vigor)
Features are aggregated according to cocoa phenology and joined with yearly FAOSTAT yield statistics.
A **Random Forest regression** model was trained and validated using a **time-based split**:
- **Train:** 2000–2018
- **Test:** 2019–2023
---
## ⚙️ Methodology
1. **Labels (FAOSTAT)**
- Download production (tonnes) & harvested area (ha).
- Compute yield = Production / Area.
2. **Seasonal Predictors**
- Aggregate CHIRPS rainfall (May–Aug sum).
- Aggregate ERA5 temperature (May–Aug mean, °C).
- Aggregate MODIS NDVI (Aug–Oct mean, scaled 0–1).
3. **Preprocessing**
- Reduce to national averages for Ghana.
- Join yearly features with yield labels.
4. **Model Training & Validation**
- Train Random Forest regression (500 trees).
- Evaluate on test years (2019–2023).
---
## 📊 Results
- **R² (Test):** ~0.62
- **RMSE / MAE:** Reasonable error margins on test years.
- **Directional Hit-Rate:** Correctly predicted yield trend (up/down) in most years.
- **Feature Importance:** Rainfall > NDVI > Temperature.
**Charts produced:**
- Time series: Observed vs Predicted yield (2019–2023).
- Scatter plot: Observed vs Predicted yields with regression line.
---
## 🚀 How to Run
### A) Run in Google Earth Engine (recommended)
1. Upload FAOSTAT Ghana cocoa CSV as a **Table asset**.
2. Open `ghana_cocoa_yield_gee.js` in GEE Code Editor.
3. Replace `FAOSTAT_ASSET` with your asset ID.
4. Click **Run** → view metrics & charts in Console.
5. In **Tasks**, export predictions CSV to Google Drive.
### B) Optional: Python Post-Processing
- Place exported CSV in `data/`.
- Install requir …
Similar
mattaaaaaswswsw/Gridded-Labour-Market-Data-for-Ghana-Using-Remote-Sensing-and-Random-ForestRandom Forest Regression in Maize Yield PredictionRandom Forest Prediction Algorithm for Eastern Africa, 1991-2023.Random forest-based mineral prospectivity modelling over the Southern Kibi–Winneba belt of Ghana using geophysical and remote sensing techniquesIntegrating Remote Sensing and Machine Learning for Crop Classification in Latur Using Google Earth Engine and Random ForestRice Yield Prediction using Machine Learning and Remote Sensing Vegetation Indices from Sentinel2, Landsat and MODIS in Mali