Logo Lanfrica

Ismailea4/morocco-weather-nowcasting

Domain:

climate

Record type:

project
Creator:
Ism
Host:
A deep learning project for short-term weather forecasting (0-4 hours) over Northern Morocco using satellite imagery from Meteosat SEVIRI and wind data from EUMETSAT HRW. # Morocco Weather Nowcasting A deep learning project for short-term weather forecasting (0-4 hours) over Northern Morocco using satellite imagery from Meteosat SEVIRI and wind data from EUMETSAT HRW. ## 🌤️ Project Overview This project implements state-of-the-art deep learning models for **weather nowcasting** - predicting immediate future weather conditions from current observations. We combine: - **Satellite imagery**: Meteosat Second Generation (MSG/SEVIRI) multi-channel data (IR, VIS, Water Vapor) - **Wind field data**: EUMETSAT High Resolution Winds (HRW) atmospheric motion vectors - **Region of Interest**: Northern Morocco (lat 21-36°N, lon -17 to -1°E) - **Temporal resolution**: 15-minute cadence predictions ### Models 1. **Baseline**: ConvLSTM encoder-decoder for spatiotemporal forecasting 2. **Advanced**: Vision Transformer (ViT) with temporal fusion for enhanced spatial pattern recognition ### Key Features - End-to-end pipeline from raw satellite data to predictions - Multi-channel weather data fusion (satellite + wind) - Comprehensive evaluation metrics (RMSE, MAE, SSIM, CSI, POD, FAR) - Visualization of predictions and attention maps - Reproducible experiments with configuration management - **🤖 AI Weather Agent**: Interactive LLM-powered assistant using ReAct pattern for real-time weather queries ## 👥 Team Structure This project is organized into three specialized roles: ### 1. Data Engineer **Responsibilities**: Data acquisition, preprocessing, and pipeline management - EUMETSAT data ingestion (SEVIRI & HRW) - Satellite image preprocessing with Satpy - Wind field gridding from BUFR format - Temporal alignment and dataset creation 📋 See detailed deliverables ### 2. ConvLSTM Baseline Engineer **Responsibilities**: Baseline model development and evaluation - Dataset loader implementation - ConvLSTM architecture design and training - Comprehensive testing and metrics - Visualization and experiment tracking 📋 See detailed deliverables ### …