Logo Lanfrica
  • Home
  • Atlas
  • Insights
  • Docs
  • Sign in

© 2026 Lanfrica. All rights reserved. All copyrights of the resources shown on the Lanfrica website belong to the original copyright holders, unless explicitly stated otherwise.

OunissiDhiaEddine/AuresNet-DZ

Domain:

climate

Record type:

model
Creator:
Oun
Host:
Bias-correction / post-processing model for GFS outputs over the Aures region (North-East Algeria), using ERA5 as target truth. # AuresNet-DZ: AI-Enhanced Weather Downscaling for Algeria A deep learning framework to bias-correct and downscale Global Forecast System (GFS) outputs (0.25°) to high-resolution ERA5-like accuracy (0.1°) specifically for the Aures mountain range in North-East Algeria. ## Core Objective The model learns a mapping to correct systematic biases in GFS caused by complex orography: $$ f(\text{GFS}) \approx \text{ERA5} $$ ## Project Structure - src/auresnet_dz/: Core Python package containing models, data modules, and training logic. - configs/: Hydra-based configuration system for models, datasets, and training loops. - scripts/: Utility scripts for data preparation and analysis. - analysis_results/: Generated metrics, error maps, and comparison plots. - checkpoints/: Model weight files (`.ckpt`). ## Quick Start ### 1. Environment Setup ```powershell python -m venv .venv .venv\Scripts\activate pip install -e . ``` ### 2. Data Preparation Align GFS and ERA5 data to the Aures grid: ```powershell python scripts/prepare_aures_data.py ``` ### 3. Execution - **Training**: ```powershell python -m auresnet_dz.train.train train.max_epochs=100 ``` - **Analysis & Reporting**: Run the inference and generate the HTML dashboard: ```powershell python scripts/generate_analysis.py --date "2023-01-18" --ckpt checkpoints/last.ckpt ``` - **View Dashboard**: Open analysis_report.html in your browser. ## Performance Analysis Suite The project includes a comprehensive analysis suite that generates: - **Weather App Dashboard**: Real-time comparison of GFS vs AI vs Truth in `analysis_report.html`. - **Improvement Metrics**: Automatically calculates % error reduction (RMSE, MAE, Bias). - **Error Maps**: Visualizes exactly where the model improves over the baseline (e.g., in high-altitude zones). ## Stack - **Framework**: PyTorch Lightning - **Architecture**: SMP U-Net (ResNet backbone) - **Data**: Xarray, Dask, NetCDF4 - **Config**: Hydra

Visit

github.com

Licenses

GPL-3.0

Similar

Hybridemotion-dz/hybridemotion-dzDZ-Digital/el-kholasa-dzSIREP-OASIS-NEXUS-TECH-DZ/AGRI_BALANCE-DZDz-EmotionDucuvuv/darija-dzMokhboud/resto-dz

Hybridemotion-dz/hybridemotion-dz

HybridEmotion DZ is an AI-powered app that analyzes hybrid Algerian language (Darija + French). It d

DZ-Digital/el-kholasa-dz

Algerian News Filtering Project with AI # الخلاصة الجزائرية | El-Kholasa DZ > **Production-Grade A

SIREP-OASIS-NEXUS-TECH-DZ/AGRI_BALANCE-DZ

AGRI_BALANCE-DZ is an Algerian smart oasis management platform built for desert agriculture across t

Dz-Emotion

Dz-Emotion is the first large-scale, manually annotated dataset for emotion detection in Algerian Ar

Ducuvuv/darija-dz

Darija algerienne — flash, QCM, decks (mobile) # Darija algérienne — apprendre pour retenir Corp

Mokhboud/resto-dz

🇩🇿 Algerian Restaurant Discovery Platform \# 🇩🇿 Resto DZ \*\*"Discover. Taste. Rate. Trust."\*\*