A high-resolution database of historical and future climate for Africa developed with deep neural networks

TL;DR

This paper introduces a detailed climate database for Africa using deep learning, covering historical and future climate data at high resolution.

Contribution

A novel deep learning approach is used to model climate data at high resolution for Africa, including historical and projected data.

Findings

The database includes climate data from 1901 to the present and CMIP6 projections for the 21st century.

A three-step method combining interpolation, deep learning, and downscaling was used to generate high-resolution climate grids.

The database is accessible via a software package and includes over 25,000 climate grids.

Abstract

This study contributes an accessible, comprehensive database of interpolated climate data for Africa that includes monthly, annual, decadal, and 30-year normal climate data for the last 120 years (1901 to present) as well as multi-model CMIP6 climate change projections for the 21st century. The database includes variables relevant for ecological research and infrastructure planning, and it comprises more than 25,000 climate grids that can be queried with a provided ClimateAF software package. In addition, 30 arcsecond (~1 km) resolution gridded data are available for download. The climate grids were developed with a three-step approach, using thin-plate spline interpolations of weather station data as a first approximation. Subsequently, a novel deep learning approach is used to model orographic precipitation, rain shadows, lake and coastal effects at moderate resolution. Lastly, lapse-rate based downscaling is applied to generate high-resolution grids. The climate estimates were optimized and cross-validated with a checkerboard approach to ensure that training data was spatially distanced from validation data. We conclude with a discussion of applications and limitations of this database.

Figures