2nd place winner — Analyzed 20 years of rainfall data to predict drought impacts on Sicilian infrastructure.
System Architecture
The project involved a rigorous data science workflow focused on SIAS hourly rainfall data (2002–2023).
- ETL Pipeline: Automated scripts to parse raw hourly precipitation logs from regional weather stations.
- Analysis Engine: Statistical modules to calculate peak intensity duration and seasonal distribution shifts.
- Projection Model: Linear extrapolation logic verified against historical trends to estimate 2050 scenarios.
Technical Challenges
Handling High-Frequency Time Series
Processing hourly data for 20 years across multiple stations resulted in a massive dataset. We optimized our Pandas operations using vectorized functions instead of loops, reducing processing time from minutes to seconds.
Correlating Disparate Metrics
The core challenge was linking rainfall deficits (meteorological data) to real-world infrastructure stress (engineering data). We created a “stress index” that normalized both datasets, allowing us to visualize exactly when and where water infrastructure would fail under predicted drought conditions.
Key Findings
Our analysis revealed a shift toward extremes: wetter summers and drier winters, with torrential events becoming shorter but more intense. This creates a “double hazard” where water is scarce when needed, but overwhelming during storms, overwhelming current storage capacities.