26/05/2025
Crop Yield Estimation of Palm Trees In AdaPalm - Nigeria’s Largest Palm Plantation.
Life makes more sense in retrospect.
A few years ago, in Owerri, I worked on a project that demanded both grit and patience. We had been commissioned to evaluate a large oil palm plantation. With drones flying overhead and boots on the ground, we painstakingly manually labelled and counted thousands of palm trees — carefully distinguishing them from other cash crops on the same land. That was around 2018/2019, and although we delivered results the client was pleased with, I couldn't shake the feeling that there had to be a better way.
Fast forward to today, after sitting through several lectures in Neural Networks and Deep Learning, something clicked. I realized that the challenges we faced back then could now be addressed with elegant, data-driven methods. I decided to relive that project — this time armed with the knowledge of remote sensing, Python, and machine learning.
One name stood out in my mind: Adapalm — the largest palm plantation in Nigeria.
What if I could estimate Adapalm’s current crop yield using satellite data? What if I could provide insight into the plantation’s recent decline using NDVI (Normalized Difference Vegetation Index) and simple yield modelling?
Setting Up the Project
I started by defining the boundary of Adapalm using a GeoJSON file. With the Copernicus Open Access Hub, I downloaded Sentinel-2 imagery, specifically Bands 4 (Red) and 8 (NIR). Then I built a Python analysis pipeline using Jupyter Notebook to compute NDVI and interpret the results.
Key Findings
AOI Area:
5,154.06 hectares — defined after reprojecting the boundary to UTM Zone 32N.
Estimated Number of Palm Trees:
737,031 trees
Based on an assumed spacing of 7 m × 7 m (49 m²/tree)
Theoretical max: ~1,052,869 trees
Effective vegetated area (NDVI > 0.35): ~13.18 million m²
Adjusted count accounts for gaps, roads, and non-productive zones
Estimated Yield:
Total yield: 6,688.5 tons
Average yield: 1.3 tons/ha
Model used:
Yield=18×(NDVI−0.35)(for NDVI > 0.35)\text{Yield} = 18 × (\text{NDVI} - 0.35) \quad \text{(for NDVI > 0.35)}
This modest yield likely reflects stress, uneven maturity, or data limitations — but it’s a solid starting point for further analysis.
Estimated Crop Value:
₦2.01 billion total, using a market price of ₦300,000/ton
₦389,312 per hectare on average
What I Learned
NDVI analysis, though simple, is powerful when paired with domain knowledge.
Satellite data can eliminate tedious fieldwork, and enable nationwide monitoring.
A few lines of Python can unlock real economic insights — from tree count to yield and value.
Why This Matters
Adapalm and many agricultural establishments in Nigeria face complex challenges — from land degradation to aging trees. Tools like this can support precision agriculture, help governments and investors track performance, and assist farmers in making data-informed decisions.
What was once a manual, field-heavy task is now a few clicks and lines of code away — thanks to free data, open tools, and a curious mind.
To my former teammates back in Owerri, I see now how far we’ve come.
To fellow GIS and remote sensing professionals — don’t underestimate the power of revisiting old problems with new skills.
Let’s keep building, learning, and applying.
