Accelerating atmospheric models using GPUs

Environmental models are simplified representations of an object or a process [1]. These models provide valuable information on the nature of real-world phenomena and systems [2], with many applications in science and engineering [3]. For example, environmental models play an increasingly important role in understanding the potential implications of climate change [4]. There are many types of models in the environmental sciences [5]. These models are often associated with large computational costs because of their complexity [6]. The model studied in this work, the Multiscale Online Nonhydrostatic AtmospheRe CHemistry model (MONARCH), is an atmospheric model that currently runs in the MareNostrum supercomputer of the Barcelona Supercomputing Center (BSC), one of the Top-500 supercomputers in the world [7] [8]. MONARCH provides regional mineral dust forecasts to the World Meteorological Organization’s (WMO) Barcelona Dust Forecast Center (BDFC) and the Sand and Dust Storm Warning Advisory and Assessment System (SDS-WAS). MONARCH also provides global aerosol forecasts to the International Cooperative for Aerosol Prediction (ICAP) initiative