Thunderstorm Prediction using Machiine Learning
- By finding correlations between the atmospheric conditions at the time of the lightning strike, we were able to predict the occurrence of lightning and thunderstorms.
What it Does
Historical data was collected from open sources for wind-speed, wind direction, temperature, precipitation, and humidity from Northern and Southern India and parts of the USA. The data was used to create a predictive model, which predicts whether a dangerous lightning strike will occur within the next 3 hours.
How We Built It
We focused on 12 districts in Northern India, with a high frequency of lightning strikes. Using data from global weather agencies and Lightning Imaging Sensor (LIS), we extracted relevant data points such as temperature, humidity, wind speed, and lightning strike times. We used an SVM classification model, which showed the highest accuracy.
Challenges We Ran Into
Finding clean and relevant datasets across the same time frame was challenging. Moreover, the dataset was imbalanced, with dangerous lightning strikes accounting for only 0.03%. We started by working on a single district and gradually built the final model.
Accomplishments That We're Proud Of
The data collection and cleaning process were time-consuming, but after extensive research, we developed a generic model that can be used for lightning prediction in various regions. We validated the model in Florida, San Diego, Kerala, and several districts in Northern India.
What We Learned
The key challenge was to find a balance between accuracy and the intensity of lightning strikes. We also learned to work with AWS SageMaker Studio Lab, which made it easy to run machine learning models.
What's Next for Thunderstorm Prediction Using ML
We hope to collaborate with global agencies for disaster response. The model can also be scaled to other geographies and disaster contexts, potentially saving many more lives.
References
- Global Hydrology Resource Center (2022). LIS Space Time Domain Search
- Holle, R. L. (2008). Annual rates of lightning fatalities by country. 20th International Lightning Detection Conference, Arizona, USA.
- Indian Meteorological Department (2022). IMD
- Kerala State Disaster Management Authority (2022). Kerala State Disaster Management Authority
- Shi, Z., et al. (2018). Effects of relative humidity on electrification and lightning discharges in thunderstorms. Terrestrial, Atmospheric & Oceanic Sciences, 29 (6).
Grey Zone Warfare Hackathon
The Grey Zone Warfare Hackathon is a dynamic event that spans over a month. This hackathon breaks away from the traditional single-problem approach by offering an open-ended format where teams can either select from a variety of suggested topics or propose their own innovative ideas.
Event overview
Participants are invited to address challenges related to grey zone warfare, focusing on practical and creative solutions in areas such as misinformation, cyber threats, emerging technologies, and national security. Some of the suggested themes include:
- Countering Misinformation: Using AI tools to combat misinformation and deepfakes.
- Stealth Drone Design: Creating drones capable of high-altitude operations.
- Cybersecurity Solutions: Developing systems for real-time threat detection.
- Simulation Models: Building models for grey zone scenarios.
- Multi-domain Integration: Designing systems that integrate various operational domains.
Participation details:
Teams must prepare an initial presentation (PPT) that outlines their chosen topic or problem statement. The presentation should include:
- Team name and details of all members.
- A clear statement of the selected topic or proposed problem.
- A concise description of the problem and potential solutions is presented in 2 to 5 slides.
Submissions will be evaluated based on the relevance of the problem, the complexity and practicality of the solution, the choice and application of technology, and the quality of the overall presentation and demonstration.
resources and tools
Participants are expected to source their own data using publicly available platforms such as Google or GenAi, or alternatively, create dummy datasets. If your project requires additional resources or specific tools, you can mention these requirements in your PPT.
Teams are encouraged to explore multiple ideas initially and then narrow them down to one after receiving feedback. For projects involving drone-based solutions, note that physical hardware is not mandatory; simulations or models are acceptable. Additionally, the focus is on developing straightforward, deployable solutions, so specialized tools like IBM Qiskit for quantum computing simulations will not be provided.
Judging and awards
The evaluation panel for the hackathon comprises experts from academia, the Indian Army, IBM, and other notable organizations. Winning teams will be recognized with awards for the first, second, and third positions, in addition to special certifications and potential extra accolades based on the judges’ assessments.
Bringing Change to Life: A Collection of Inspiring Photos
This website uses cookies.
We use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, your data will be aggregated with all other user data.