Data and Methods
Climate Predictions in Sudan
Exploring temperature and rainfall trends to support
climate understanding using data science
Weather shifts shape the way communities live, grow and plan their future.
Yet many regions still face unpredictable rainfall and rising temperatures
that threaten farming, water stability and daily life. Behind these changes
are patterns that are rarely measured in Sudan.
This project studies these patterns through clean data and scientific
analysis, helping reveal how rainfall and temperature behave over time and
making predictions that support better planning and understanding
This project uses the NASA POWER database to collect climate data for Sudan, focusing on 2-meter air temperature (°C) and total corrected precipitation (mm/day).
Sudan is divided into five regions for analysis: North, Central, East, West, and South. Data for each region are collected, cleaned, and merged into structured datasets suitable for machine learning modeling.
Only temperature and rainfall are included in this version, while other climate variables like humidity, wind, or pressure could improve accuracy in future iterations.
Project Workflow
Learn More
Findings & Insights
Findings
Rainfall is more irregular, so predictions vary between regions and some months are harder to forecast accurately.
Temperature follows a steady seasonal pattern, making predictions relatively reliable.
Forecasts are intended as guidance to understand trends, rather than exact predictions for operational planning.
Forecasting Dataset
View Full Analysis
Regional Insights
Temperature trends: Temperatures are slightly increasing in recent years.
Rainfall variability: Southern and Western regions show more fluctuation, while the North remains mostly dry and stable.
Inverse relationships: Most regions (Central, East, South, West) show negative correlations between rainfall and temperature.
Northern region anomaly: Slight positive correlation due to very low rainfall.
Future Work and Improvements
This project is only the first step toward building a reliable climate insight tool. While the current version provides structured data, clear trends and early predictive models, there are several areas that can be improved.
Goals for Future Development
Expand the dataset
Include more climate variables, such as humidity, wind speed and soil moisture. This will give a broader view of environmental conditions.Improve the prediction models
Test more advanced machine learning methods and tune the existing models to reduce error and improve long term reliability.Add interactive features
Integrate map based visualizations, real time updates and a simple user interface so anyone can explore climate trends directly.Increase geographical detail
Collect data at the city or village level instead of coordinate-based points. This makes results easier to interpret for real communities.
Current Limitations
Only rainfall and temperature are included, which limits the ability to capture full climate relationships.
The dataset does not cover local administrative boundaries, which reduces its direct usefulness for decision making.
Predictions become less accurate during months with high variation, especially with rainfall values.
These issues guide what needs improvement and help shape the next phase of the project.
If you are interested in this work, have suggestions or want to collaborate, feel free to reach out. I am always open to new ideas that can support stronger and more reliable climate insights for communities.
Developed individually by Mohamed Tilal as part of the MIT Emerging Talent program (ELO2).
GitHub