Climate-Smart Agriculture

Problem Statement

Smallholder farmers in East Africa face increasing climate uncertainty — erratic rainfall, rising temperatures, and shifting seasons make traditional farming knowledge less reliable. This project analysed crop yield data against climate variables to identify patterns that could help farmers and agricultural planners make better decisions about what to grow, when to plant, and where to invest.

Approach

Starting from a raw Jupyter notebook, I built a structured Python pipeline that takes agricultural data from ingestion through to final insights. The pipeline was designed to be reproducible and transparent — each stage clearly documented so that results can be verified and trusted.

Loaded and validated raw agricultural datasets using Pandas
Cleaned missing values, corrected data types, and removed outliers
Performed statistical analysis to identify correlations between climate variables and crop yields
Built visualisations with Matplotlib to communicate findings clearly
Exported a summarised Excel report for non-technical stakeholders

Pipeline Architecture

The pipeline follows a clean 4-stage structure:

# Stage 1: Data Ingestion
import pandas as pd
df = pd.read_csv('agriculture_data.csv')

# Stage 2: Data Cleaning
df.dropna(subset=['yield_kg', 'rainfall_mm'], inplace=True)
df['season'] = df['date'].dt.quarter

# Stage 3: Statistical Analysis
correlation = df[['yield_kg', 'rainfall_mm', 'temp_avg']].corr()

# Stage 4: Export Summary
df.to_excel('agriculture_summary.xlsx', index=False)
        

Screenshots

Excel summary report — Excel Summary — Crop yield analysis for non-technical stakeholders

Python data analysis visualisations — Python — Statistical analysis and visualisations (Matplotlib)

Python data pipeline — Python — Data pipeline in Jupyter Notebook

Results

Identified strong correlation between seasonal rainfall and maize yields across regions
Flagged temperature anomaly years where yields dropped significantly
Produced a clean Excel summary accessible to agricultural extension officers
Built a reproducible pipeline that can be rerun as new data becomes available

Skills Demonstrated

Python data pipeline design (Pandas, Matplotlib)
Data cleaning and validation at scale
Statistical correlation analysis
Data visualisation for storytelling
Excel reporting for non-technical audiences
Jupyter Notebook documentation