Climate risk means different things to different people. For a farmer in Nsanje it means drought. For a planner in Blantyre it means flooding. For a policy maker in Lilongwe it means both, happening in different places at the same time. The challenge is not understanding that risk exists. It is quantifying it in a way that supports decisions.

This project was my attempt to do that systematically across all 28 Malawian districts.

The Framework

I applied the IPCC AR5 risk framework, which defines risk as a function of three components: hazard, exposure, and vulnerability. Rather than treating these as abstract concepts, I operationalized each one using measurable climate indices derived from daily meteorological data.

The primary indices used were the Standardized Precipitation Index (SPI) for drought hazard and the Coefficient of Variation for rainfall variability. Normalization used 5th to 95th percentile clipping to handle extreme outliers without distorting district-level comparisons.

The Data

The pipeline processes over 50,000 daily meteorological records from 2020 to 2024, sourced from the NASA POWER API for all 28 districts. Each district gets its own time series, risk score, and index decomposition.

The Dashboard

The Streamlit dashboard uses Plotly choropleth maps to visualize risk scores geographically, with radar charts showing the breakdown of hazard, exposure, and vulnerability components per district. Policymakers can identify high-risk districts in under five seconds. The interface is mobile-optimized.

The goal was not to build a research tool. It was to build something a non-technical decision maker could open and understand immediately.

What I Learned

Working with district-level aggregation taught me that spatial scale matters enormously. A national average hides everything. The districts with the highest risk scores in Malawi are not the ones most people would guess, and that gap between assumption and data is exactly where this kind of tool becomes useful.

Live Demo →