Saved in:
| Main Author: | |
|---|---|
| Format: | Recurso digital |
| Language: | |
| Published: |
Zenodo
2026
|
| Online Access: | https://doi.org/10.5281/zenodo.19394419 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Table of Contents:
- <p>This dataset contains data and analysis code used to generate the results presented in our paper "Interpretable machine learning reveals context-dependent decision-making in migratory songbirds". It is structured as follows:</p> <ul> <li><strong>data</strong>: preproceseed datasets used to train, test, and analyse FluxRGNN+ bird migration models <ul> <li><em>preprocessed</em>: hourly NEXRAD weather radar data combined with ERA5 reanalysis data and H3 tessellation details</li> <li><em>shapes: </em>geographical shapes used for plotting</li> </ul> </li> <li><strong>raw</strong>: raw NEXRAD weather radar data used to generate the preprocessed datasets. <ul> <li><em>nexrad</em>: vertically integrated bird densities and average velocities used to train and evaluate FluxRGNN+ models</li> <li><em>nexrad_vp_avg</em>: average vertical profiles per radar and season used to determine ERA5 model levels</li> </ul> </li> <li><strong>analysis</strong><em>: </em>all relevant data and<em> </em>code used for the final analysis of FluxRGNN+ predictions. The <em>environment.yml </em>file can be used to install all required Python packages with conda. <ul> <li><em>artifacts: </em>trained models, predictions, and Shapley values. The <em>.csv</em> files in the <em>analysis</em> folder can be used to find specific files in this folder (e.g. the file <em>final_model_artifacts.csv</em> contains the path to the trained models for spring and fall migration seasons respectively)</li> <li><em>shap_values: </em>processed (filtered and grouped) Shapley values generated by the script <em>process_explanations.py</em></li> <li><em>plots: </em>figures generated by the Jupyter notebooks <em>evaluate_models.ipynb, plot_predictions.ipynb, </em>and <em>analyse_explanations.ipynb</em></li> </ul> </li> </ul> <p>Model training and Shapley value computations were performed using the publically available <a href="https://github.com/FionaLippert/FluxRGNN" target="_blank" rel="noopener">FluxRGNN</a> (v2.0.0) and <a href="https://github.com/shap/shap" target="_blank" rel="noopener">shap</a> (v0.45.1) Github repositories.</p>