A modern desktop weather app that asks four world-class models, shows you where they disagree, and grades its own accuracy.

✨ Features  •  📸 Screenshots  •  🚀 Quick Start  •  🔬 How It Works  •  🗺️ Roadmap

🌟 At a Glance

4 weather models queried in parallel

24h + 5d hourly strip and daily outlook

0 API keys required

MAE + Hit Rate self-scored vs. persistence & IFS

---

🎯 Why This Project

Most weather apps present one model's output as if it were certain. But models disagree, and that disagreement is useful information.

Weather Forecasting queries several independent models at once, reports the median, shows how far apart they are, and then checks itself against what actually happened, so you can see how much to trust a forecast, not just what it says.

💡 Narrow spread = models agree. Wide spread = treat the forecast with caution.

---

✨ Features

🧠 Ensemble Forecasts Queries ECMWF IFS, ECMWF AIFS, GFS and ICON in parallel and reports the median temperature with a confidence spread.	📈 Self-Scoring Accuracy Grades forecasts against observed conditions and benchmarks them against persistence and IFS baselines.	🚨 Severe Weather Alerts Automatic detection of heat, cold, wind, rain and storm conditions.
📊 Location Scoring Ranks locations by comfort, safety and overall conditions.	🌗 Sun & Moon Arc Daylight curve with sunrise and sunset markers, plus the current moon phase.	📅 Hourly & 5-Day A 24-hour strip and 5-day outlook with precipitation probabilities.
🎨 Adaptive Dark UI Card-based layout with dynamic colors and icons that follow conditions.	🔍 Search & Geolocation City search with geolocation calibration.	🖥️ Headless CLI A quick terminal preview of the full ensemble, great for scripting.

---

📸 Screenshots

⏱️ Hourly strip · 5-day outlook · Forecast accuracy

⌨️ CLI preview · Multi-model ensemble output

---

🔬 How It Works

flowchart LR
    A["City search<br/>or geolocation"] --> B["fetcher.py"]
    B --> C1["ECMWF IFS"]
    B --> C2["ECMWF AIFS"]
    B --> C3["GFS"]
    B --> C4["ICON"]
    C1 --> D["Ensemble merge<br/>median + spread"]
    C2 --> D
    C3 --> D
    C4 --> D
    D --> E["alerts.py"]
    D --> F["scoring.py"]
    D --> G["ui.py + widgets.py"]
    E --> G
    F --> G

    style A fill:#1b5fa8,color:#fff
    style D fill:#8b5cf6,color:#fff
    style G fill:#22c55e,color:#fff

Loading

Step	What happens
1 · Fetch	Open-Meteo is queried once per configured model, in parallel
2 · Merge	Responses are normalized; the median is reported and the spread becomes the confidence signal
3 · Assess	Alert rules flag hazards; scoring ranks comfort and safety
4 · Grade	Past forecasts are compared with observations using MAE and hit rate, against persistence and IFS baselines

---

🚀 Quick Start

Requirements: Python 3.10+ and an internet connection. No API key.

1 · Clone

git clone https://github.com/sonubiswal/Weather-Forecasting.git
cd Weather-Forecasting

2 · Create a virtual environment

🪟 Windows	🍎 macOS / 🐧 Linux
python -m venv venv venv\Scripts\activate	python3 -m venv venv source venv/bin/activate

3 · Install

pip install -r requirements.txt

4 · Run

🖥️ Desktop GUI	⌨️ CLI preview
cd src python -m weather_app.main	python preview.py

📟 Sample CLI output

Puri, India  (19.798, 85.825)
  median temperature : 26.1 °C  (spread 2.1° across 4/4 models)
  feels like         : 31.6 °C
  humidity / wind    : 88% / 4.8 km/h
  condition          : Clear sky
  confidence         : Good confidence · 4 models within 2.1°
  hours / days       : 24 / 5
  models configured  : ecmwf_ifs025, ecmwf_aifs025_single, gfs_seamless, icon_seamless

---

🏗️ Architecture

Weather-Forecasting/
├── 🖼️  assets/
│   ├── banner.svg
│   └── screenshots/
├── 📦 src/weather_app/
│   ├── __init__.py     # Package metadata and version
│   ├── main.py         # Entry point (QApplication + event loop)
│   ├── fetcher.py      # Open-Meteo client, ensemble merge, normalization
│   ├── scoring.py      # Comfort / safety scoring, accuracy tracking
│   ├── alerts.py       # Severe weather detection rules
│   ├── theme.py        # Dynamic color, font, and icon selection
│   ├── widgets.py      # Reusable PySide6 components (cards, strips, arcs)
│   └── ui.py           # Main window layout and event wiring
├── preview.py          # Headless CLI preview
├── requirements.txt
├── pyproject.toml
└── .env.example

Principle	In practice
Separation of concerns	Data (fetcher), logic (scoring, alerts) and presentation (ui, widgets, theme) are separate modules
Reusable widgets	UI building blocks are isolated in widgets.py
Provider-agnostic data layer	Normalization in fetcher.py keeps the rest of the app independent of the API's response shape

---

⚙️ Configuration

Open-Meteo needs no key, so this is optional. To add a custom provider later:

🪟 Windows	🍎 macOS / 🐧 Linux
copy .env.example .env	cp .env.example .env

Model ID	Provider
ecmwf_ifs025	🇪🇺 ECMWF IFS
ecmwf_aifs025_single	🤖 ECMWF AIFS
gfs_seamless	🇺🇸 NOAA GFS
icon_seamless	🇩🇪 DWD ICON

---

🩺 Troubleshooting

❌ ModuleNotFoundError: No module named 'weather_app'

Run from inside src using the module form:

cd src
python -m weather_app.main

🌐 The app opens but shows no data

Check your internet connection and that Open-Meteo is reachable. The CLI line shows how many models responded (for example 4/4 models).

📦 PySide6 fails to install or the window won't launch

Use Python 3.10+ and upgrade pip first:

python -m pip install --upgrade pip
pip install -r requirements.txt

On some Linux distributions, Qt may need extra system libraries (for example libxcb-related packages).

🪟 venv won't activate in PowerShell

Use Command Prompt, or allow scripts for the current session:

Set-ExecutionPolicy -Scope Process -ExecutionPolicy RemoteSigned

---

🛠️ Tech Stack

---

🗺️ Roadmap

• 📉 Historical trend charts (30 / 90-day temperature and rainfall)
• ⭐ Saved locations and quick-switch favorites
• 🔔 System tray notifications for active alerts
• 🌓 Light / dark theme toggle
• 📦 Standalone executable (PyInstaller)
• 🏆 Expanded accuracy dashboard (per-model scoring breakdown)

Have an idea? Open an issue 💬

---

🤝 Contributing

Contributions, bug reports and feature requests are welcome.

1. 💬 Open an issue first to discuss any major change
2. 🍴 Fork the repo and create a branch: git checkout -b feature/your-feature
3. ✍️ Commit with clear, descriptive messages
4. 🚀 Open a Pull Request explaining what changed and why

Please keep changes focused, match the existing module structure, and include screenshots for UI changes.

---

🙏 Data Attribution

Weather data by Open-Meteo.com under CC BY 4.0. Forecasts derive from models operated by ECMWF, NOAA (GFS) and DWD (ICON).

⚠️ Disclaimer: For informational use only. Do not rely on this app as your sole source for safety-critical decisions; consult official meteorological authorities during severe weather.

---

📄 License

Released under the MIT License. See LICENSE.

👤 Author

Sonu Biswal · @sonubiswal

⭐ If this project helped you, give it a star!

_{Built with Python, PySide6 and Open-Meteo}