Climate is a hot topic, and not just in the media. Researchers from around the world are working on the subject and towards improved weather forecasts and climate projections to better assess, prepare for and address climate-related risks. In this day and age, this type of research can hardly be done without appropriate and efficient research software.The Netherlands eScience Center (NLeSC) and Atos have therefore teamed up to support four new projects aiming to improve Europe’s weather and climate models with their research software expertise.  

These projects are funded as part of the open call in ESiWACE2 Service 1 published in autumn 2021. The ESiWACE2 Services enable short collaborative projects that offer guidance, engineering and advice by the ESiWACE2 consortium partners to support exascale preparations for weather and climate models in Europe. Known as Service 1 projects, they target developing communities and assist with porting codes to new hardware platforms, such as Graphics Processing Units (GPUs). Such platforms will improve model efficiency and prepare the software to enable model execution on existing and near-future hardware architectures, as well as simulate experiments at unprecedented grid resolutions or ensemble sizes. 

This is the third and final year of Service 1 projects within ESiWACE2. After two years of successfully collaborating with scientists to accelerate the software underlying climate simulations, we are excited to continue our efforts to prepare the weather and climate modelling communities in Europe for the upcoming exascale revolution. Congratulations to all the winning projects! 

The new project collaborations awarded in 2022 are:

OGS Transport model (OGSTM) 

National Institute of Oceanography and Applied Geophysics (OGS), Italy — Paolo Lazzari

OGSTM is an oceanographic transport model that is coupled with a Biogeochemical Flux Model, with the aim of modelling marine biogeochemistry. This software is routinely used as the transport component of the biogeochemical state of the Mediterranean Sea (MedBFM) forecasting system. The latter is embedded in the EU Copernicus Marine Service (CMS) to operationally produce analysis and forecasts of biogeochemistry in the Mediterranean Sea.

In this project, we will investigate how the performance of the code may be improved by using GPUs.

Regional Model: REGCM4 
The Abdus Salam International Centre for Theoretical Physics (ICTP), Italy — Graziano Giuliani

The Regional Climate Model system (RegCM) is flexible, portable and easy to use. It can be applied to any region of the world, with grid spacing of up to about 10 km (hydrostatic limit), and to a wide range of studies, from process studies to paleoclimate and future climate simulation.

In this project, we will investigate how GPUs can be employed to accelerate the compute-intensive operations in the model. The goal is to use GPUs to enable the 900+ RegCM user community to perform simulations on higher resolutions, for example 3km, in order to evaluate the future of climate extremes. These extremes are expected to change over multiple regions all over the world. A number of pilot studies are currently targeting high threat areas for climate change such as the Indo-Gangetic plane, Southern Europe, the Sahel Region and the la Plata basin. The RegCM model can provide useful information on future extremes in temperature and precipitation, serving as red flag warnings for possible life disruptive changes in different regions around the globe.

Wageningen University and Research, the Netherlands — Bart van Stratum, Chiel van Heerwaarden

MicroHH is a software for large-eddy simulation (LES), which is primarily used to study turbulent flows in the atmosphere. The main difference with numerical weather prediction codes is that LES models do not rely on physical parameterisations for turbulence and clouds. These are explicitly resolved by the LES model, which can lead to higher quality weather forecasts. MicroHH is already implemented in CUDA, a programming language specific to GPUs. In this project, we will investigate how to improve the performance of the existing CUDA implementation of MicroHH and optimise the performance of the code. We plan to use Kernel Tuner, developed by the Netherlands eScience Center, to automatically optimise the GPU code for the best performance. Finally, we will also look into accelerating MicroHH on multi-GPU systems. 

NORCE, Norway — Alok Kumar Gupta

The Bergen Layered Ocean Model (BLOM) is the ocean component of the Norwegian Earth System Model (NorESM). NorESM itself is a global Earth system model that figured in the recent CMIP5 and CMIP6 projects, and it is expected that contributions to the upcoming international intercomparison projects will make use of it. The goal of this project is to improve the model’s performance on the latest computing architectures, to enable it to run at higher resolutions for more fine-grained simulations. In this project, the Research Software Engineers at Atos and the eScience Center working within ESiWACE2 will investigate how to improve the performance of the software using OpenMP. Furthermore, they will explore the possibilities of using GPUs through directive-based approaches such as OpenMP and OpenACC

To learn more about the services offered within ESiWACE2, please visit our page on Services.