• English 
  • Spanish 

Reproducible Statistical Downscaling with the climate4R R-Based Framework: The downscaleR package

Conference: ICRC-CORDEX 2019: International Conference on Regional Climate
Year: 2019
Contribution type: Oral

climate4R (https://github.com/SantanderMetGroup/climate4R) is an R-based open framework for climate data access, harmonization and post-processing. The framework allows for comprehensive end-to-end sectoral reproducible applications through the interoperability of different specific R packages. One of the core packages is loadeR, the climate4R catalog- and subset-oriented data access tool from either local or remote (OpeNDAP) data sources, which is transparently linked to the Santander Climate Data Service (UDG), providing a wide catalogue of popular datasets (incl. CMIP and CORDEX) to climate4R users. As part of the climate4R ecosystem, the package downscaleR provides tools for bias adjustment and statistical downscaling, covering the most popular techniques (e.g. analogs, LMs and GLMs, and neural networks), and allowing for multiple experiment configurations and cross-validation options. climate4R was used to contribute to the VALUE intercomparison experiment, and will be used in ongoing experiments in the framework of EURO-CORDEX.

In order to illustrate the functionalities of downscaleR and climate4R, we present a fully reproducible worked example from the VALUE contribution, obtaining local climate change projections for a set of 86 stations over Europe. We first replicate the VALUE 1A Experiment (with reanalysis predictors) using an extended set of methods. Then, we describe ongoing work with GCM predictors to compute the climate change signal for the late XXI century and analyze the spread resulting from the ensemble of statistical downscaling methods used.

Climate data processing typically involves complex error-prone operations. In this sense, climate4R provides a unique framework where common tasks such as statistical downscaling can be straightforwardly performed in a few lines of code. The development of climate4R is a community effort boosted by the contribution to several international initiatives, such as the IPCC WGI activities of the Atlas Chapter.