Crop Calendar Climate Shift Analyser 🌧️🌽

Has Malawi's planting season shifted since 1990?

A data science project using ERA5 reanalysis and CHIRPS rainfall data to detect shifts in onset-of-rains (OOR) dates across Malawi's districts from 1990–2024, with implications for smallholder maize farming.

---

Research Question

Climate change is altering rainfall seasonality across sub-Saharan Africa. This project quantifies how much and where Malawi's agricultural calendar has shifted by:

1. Extracting onset-of-rains (OOR) dates per district (1990–2024)
2. Fitting linear trends to detect statistically significant shifts
3. Mapping spatial patterns of change
4. Generating an interactive Streamlit dashboard and Quarto report

---

Project Structure

crop-calendar-shift/
│
├── data/
│   ├── raw/                  # Downloaded ERA5 / CHIRPS NetCDF files
│   └── processed/            # Cleaned CSVs ready for analysis
│
├── scripts/
│   ├── python/
│   │   ├── 01_download_chirps.py     # Download CHIRPS rainfall via API
│   │   ├── 02_extract_oor_dates.py   # Compute onset-of-rains per district
│   │   └── 03_export_for_r.py        # Export clean CSV for R analysis
│   └── r/
│       ├── 04_trend_analysis.R       # Linear trend fitting + Mann-Kendall
│       └── 05_visualisations.R       # ggplot2 maps and charts → outputs/
│
├── dashboard/
│   └── app.py                # Streamlit interactive dashboard
│
├── report/
│   ├── report.qmd            # Quarto report (rendered to HTML/PDF)
│   └── references.bib
│
├── outputs/
│   ├── figures/              # All saved plots
│   └── tables/               # Trend result tables (CSV)
│
├── environment.yml           # Conda environment
├── requirements.txt          # pip requirements
├── renv.lock                 # R package lockfile (generated by renv)
└── README.md

---

Quickstart

1. Set up Python environment

conda env create -f environment.yml
conda activate crop-calendar

Or with pip:

pip install -r requirements.txt

2. Download rainfall data

python scripts/python/01_download_chirps.py
python scripts/python/02_extract_oor_dates.py
python scripts/python/03_export_for_r.py

3. Run R analysis

source("scripts/r/04_trend_analysis.R")
source("scripts/r/05_visualisations.R")

4. Launch dashboard

streamlit run dashboard/app.py

5. Render Quarto report

quarto render report/report.qmd

---

Methodology

Onset-of-Rains Definition

Following FAO/FEWS NET standards, OOR is defined as the first day after October 1st when:

• Cumulative rainfall ≥ 25mm over 3 consecutive days
• No dry spell > 10 days in the following 20 days

Trend Detection

• Sen's slope estimator — robust linear trend (mm/year or days/decade)
• Mann-Kendall test — non-parametric significance test (p < 0.05)
• Applied per district, significance mapped spatially

---

Data Sources

Dataset	Variable	Resolution	Source
CHIRPS v2.0	Daily precipitation	0.05°	UCSB CHG
ERA5-Land	2m temperature, ET	0.1°	Copernicus CDS
GADM Malawi	District boundaries	Vector	GADM

---

Author

George Sichinga
MSc Applied Data Science (Environment & Climate Modelling)
LUANAR, Malawi
georgesichinga.github.io

---

License

MIT