Transparency and reproducibility are key ingredients to develop top-quality science. For this reason, this repository is aimed at hosting and maintaining updated versions of the code and notebooks needed to (partly or fully) reproduce the results of the papers developed in the Santander MetGroup dealing with the application of deep learning techniques for statistical dowscaling in climate.

These works build on climate4R, a bundle of R packages developed for transparent climate data access, post processing (including bias correction and downscaling), visualization and model validation. A battery of Jupyter notebooks with worked examples explaining how to use the main functionalities of the core climate4R packages (including downscaleR for standard statistical downscaling methods) can be found at the notebooks' repositoty. For deep learning impplementations we use keras, an R library which provides an interface to Keras, a high-level neural networks API which supports arbitrary network architectures and is seamlessly integrated with TensorFlow, and a wrapper of this package for the downscaleR package, downscaleR.keras.

The table below lists the documents (Jupyter notebooks, scripts, etc.) contained in this respository along with the information of the corresponding published (or submitted) papers.