Scikit-downscale: toolkit for statistical downscaling

Scikit-downscale is a toolkit for statistical downscaling using Scikit-Learn. It is meant to support the development of new and existing downscaling methods in a common framework. It implements Scikit-learn’s fit/predict API facilitating the development of a wide range of statistical downscaling models. Utilities and a high-level API built on Xarray and Dask support both point-wise and global downscaling applications.

Quick Start

Install scikit-downscale:

pip install scikit-downscale

Then try your first downscaling example:

from skdownscale.pointwise_models import QuantileMapper

# Initialize the model
qm = QuantileMapper()

# Fit on training data
qm.fit(model_data, observations)

# Generate downscaled predictions
downscaled = qm.predict(model_data)

Ready to learn more? Check out our Getting Started!

Documentation Structure

This documentation is organized following the Diátaxis framework:

📚 Tutorials - Learning-oriented

Step-by-step lessons to learn scikit-downscale. Start here if you’re new!

🔧 How-to Guides - Problem-oriented

Practical guides for accomplishing specific tasks.

📖 Background - Understanding-oriented

Explanations of concepts, theory, and design decisions.

📋 Reference - Information-oriented

Complete API documentation and technical specifications.

Project Info

Under Active Development

Scikit-downscale is under active development. We are looking for additional contributors to help fill out the list of downscaling methods supported here. We are also looking to find collaborators interested in using deep learning to build global downscaling tools. Get in touch with us on our GitHub page.

Indices and tables