⚡ThunderCast: Storm Forecasting Platform

Real-time Thunderstorm Prediction System using Time-Series Forecasting

---

🚀 Live Demo

Try it now: ThunderCast Live on Streamlit →

---

📋 Project Overview

ThunderCast is an automated thunderstorm prediction system that uses Facebook Prophet (time-series forecasting) to predict thunderstorm probabilities for Pimpri-Chinchwad, Pune. The system collects real-time weather data, analyzes patterns, and provides predictions with an interactive dashboard.

---

🎯 Features

• ✅ Real-time Weather Data Collection - Automated hourly data collection via OpenWeatherMap API
• ✅ Time-Series Forecasting - Facebook Prophet model for thunderstorm prediction
• ✅ Interactive Dashboard - Built with Streamlit and Plotly for data visualization
• ✅ Historical Analysis - EDA and statistical insights on 13+ years of weather data
• ✅ Automated Predictions - 6-24 hour thunderstorm probability forecasts
• ✅ Cloud Database - Supabase for data storage and retrieval

---

🛠️ Tech Stack

Languages & Libraries:

• Python 3.13
• Pandas & NumPy (Data manipulation)
• Prophet (Time-series forecasting)
• Plotly (Interactive visualizations)
• Streamlit (Dashboard framework)

Tools & Services:

• Supabase (PostgreSQL database)
• OpenWeatherMap API (Weather data source)
• APScheduler (Task automation)

Domain:

• Time-Series Forecasting
• Weather Analytics
• Predictive Analytics

---

📂 Project Structure

ThunderCast/
│
├── data/
│   ├── raw/              # Raw Kaggle dataset
│   ├── processed/        # Cleaned and featured data
│   ├── models/           # Trained Prophet model
│   └── visualizations/   # Generated charts (Plotly HTML files)
│
├── src/
│   ├── data_collection.py      # Fetch weather data from API
│   ├── data_exploration.py     # Initial data exploration
│   ├── data_cleaning.py        # Data cleaning and preprocessing
│   ├── eda_analysis.py         # Exploratory Data Analysis
│   ├── feature_engineering.py  # Feature creation
│   ├── model_training.py       # Prophet model training
│   ├── prediction.py           # Generate predictions
│   ├── scheduler.py            # Automated data collection
│   └── visualizations.py       # Create Plotly charts
│
├── dashboard/
│   └── app.py            # Streamlit dashboard
│
├── config/
│   └── config.py         # Configuration settings
│
├── .env                  # Environment variables (API keys)
├── requirements.txt      # Python dependencies
└── README.md            # Project documentation

---

🚀 Installation & Setup

Prerequisites

• Python 3.13+
• pip package manager
• Supabase account (free tier)
• OpenWeatherMap API key (free)

Step 1: Clone Repository

git clone <repository-url>
cd ThunderCast

Step 2: Create Virtual Environment

python -m venv venv
venv\Scripts\activate  # Windows
# source venv/bin/activate  # Mac/Linux

Step 3: Install Dependencies

pip install -r requirements.txt

Step 4: Configure Environment Variables

Create .env file in project root:

SUPABASE_URL=your_supabase_url
SUPABASE_KEY=your_supabase_anon_key
OPENWEATHER_API_KEY=your_openweather_api_key

Step 5: Setup Database

Run SQL script in Supabase SQL Editor to create tables:

-- weather_data table
CREATE TABLE weather_data (
    id BIGSERIAL PRIMARY KEY,
    timestamp TIMESTAMPTZ DEFAULT NOW(),
    temperature FLOAT,
    humidity FLOAT,
    pressure FLOAT,
    wind_speed FLOAT,
    cloud_cover FLOAT,
    location VARCHAR(100),
    latitude FLOAT,
    longitude FLOAT
);

-- predictions table
CREATE TABLE predictions (
    id BIGSERIAL PRIMARY KEY,
    prediction_time TIMESTAMPTZ DEFAULT NOW(),
    forecast_time TIMESTAMPTZ,
    thunderstorm_probability FLOAT,
    location VARCHAR(100),
    model_version VARCHAR(50)
);

-- alerts table (optional)
CREATE TABLE alerts (
    id BIGSERIAL PRIMARY KEY,
    alert_time TIMESTAMPTZ DEFAULT NOW(),
    location VARCHAR(100),
    severity VARCHAR(20),
    message TEXT,
    sent BOOLEAN DEFAULT FALSE
);

---

📊 Usage

1. Collect Weather Data

python src/data_collection.py

2. Data Cleaning & Feature Engineering

python src/data_cleaning.py
python src/feature_engineering.py

3. Train Model

python src/model_training.py

4. Generate Predictions

python src/prediction.py

5. Launch Dashboard

streamlit run dashboard/app.py

6. Run Automated Scheduler (Optional)

python src/scheduler.py

---

📈 Model Details

Algorithm: Facebook Prophet (Time-Series Forecasting)

Features Used:

• Temperature (°C)
• Humidity (%)
• Atmospheric Pressure (hPa)
• Wind Speed (km/h)
• Cloud Cover (%)
• Precipitation (mm)

Target Variable: Thunderstorm occurrence (Binary: 0/1)

Training Data: 116,135 hourly records (2008-2022, Pune weather)

Prediction Horizon: 6-24 hours ahead

Performance Metrics:

• Model captures seasonal patterns
• Identifies high-risk conditions (humidity >70%, pressure <1010 hPa)
• Low false positive rate for clear weather conditions

---

🎨 Dashboard Features

Live Weather Monitoring:

• Current temperature, humidity, pressure, wind speed, cloud cover

Thunderstorm Predictions:

• 6-24 hour probability forecast
• Risk level indicators (Low/Moderate/High)
• Interactive Plotly charts

Historical Trends:

• Multi-parameter weather visualizations
• Statistical summaries
• Customizable time ranges (6h, 12h, 24h, 7 days)

---

🔑 Key Insights from EDA

1. Thunderstorm Conditions:

   • Average humidity during storms: 85%+
   • Average pressure during storms: <1008 hPa
   • Peak occurrence: Monsoon season (June-September)

2. Temporal Patterns:

   • Higher frequency in afternoon/evening hours
   • Seasonal variations clearly visible

3. Data Quality:

   • 0% missing values
   • 116,136 complete records
   • 482 thunderstorm events identified (0.42%)

---

📝 License

This project is for educational purposes.

---

🙏 Acknowledgments

• OpenWeatherMap for weather data API
• Facebook (Meta) for Prophet library
• Supabase for cloud database
• Kaggle for historical weather dataset

---

📞 Contact

For queries or collaboration:

• 👨‍💻 Author Name: Akash Yadav
• GitHub: @akxyverse
• Email: akashyadav110502@gmail.com
• LinkedIn: [https://www.linkedin.com/in/akash-yadav-122a75288/]

---

Built with ❤️ using Python, Prophet, and Streamlit