Periodic Reporting for period 2 - NextGEMS (Next Generation Earth Modelling Systems)
Reporting period: 2023-05-01 to 2024-04-30
nextGEMS is a four year project with the simple objective of developing and applying a new generation of storm-resolving (km-scale) Earth system models for applications, for science, and for new communities of users. nextGEMS exploits advances in computational capacity to explicitly represent the main processes responsible for vertical energy transport in the atmosphere, and lateral transport in the ocean — processes that existing climate models leave out, or parameterize using statistical-empirical models. This approach results in a representation of the climate that evolves globally with local (km-scale) granularity, making for more tangible and salient output, thereby bridging the gap to both impacts and observations. The development of the models poses extraordinary challenges not only for Earth-system scientists, but also computational scientists. Making the computations run efficiently on modern computational infrastructures, and efficiently managing their extra-ordinary amount of output for a world-wide user and developer community poses grand challenges.
A little more than midway through the project nextGEMS has created two storm-resolving models. One is based on the ICON modelling framework, the other the Integrated Forecast System (IFS) of the European Center for Medium Range Forecasts (ECMWF). Both models have been developed to exploit the full bandwidth of high-performance computing architectures and as a result were the first (and still only) Earth system models capable of using Europe’s most performant, efficient, and carbon neutral computing facilities (LUMI and JEDI). New hierarchical, cache-optimized, methods have been demonstrated for managing output to maximize the global user experience. Through its pioneering use of Hackathons nextGEMS has demonstrated the efficiency of remote data handling, advanced best practices, and built an integrated international community of users outside of the major modelling centers.
Developing the models to run for climate applications identified many errors that had been hidden from previous generations of the models either by their exclusive use for short-timescales, or by the wide-spread practice of model tuning. By maintaining full control of the mass and energy budgets, and in particular physical processes influencing low clouds, both models could be shown to reproduce the annual budget of energy and temperature, and to run stably for periods of decades. Both models were run with a grid spacing of 10 km in the atmosphere and 5 km in the ocean for multi decadal (30 yr) projections based on a plausible emission scenario (SSP3-7.0) and the data has been published. The resolution now provides access to climate information for regions and on scales never before covered by future climate projections. Both models are being expanded to include new Earth-system related components. The fine scale of the models allows for a more physical coupling to a large variety of processes and establishes the storm-resolving modelling framework as a more physically compelling representation of the Earth system as a whole.
The models ability to exploit computational resources provided by EURO HPC has allowed the numerical experimentation required for their application to a variety of scientific questions. These include factors con- trolling the distribution of tropical precipitation, convective organization, low level stratiform clouds in the tropics, the energetic response to warming, the response to aerosol forcing, the effect of land-use changes on precipitation, and the influence of fine-scale heterogeneity on large-scale patterns of warming. Scientific findings include the important of air-sea energy exchange in low wind regimes for the distribution of tropical convection, a negative feedback between precipitation and soil moisture, and the importance of vertical mixing in regimes of stable stratification for controlling the distribution of cloud amount.
Storyline approaches have been guiding the use of output for applications with foci on renewable energy and food security. Coupled ocean-biogeochemical simulations enable a first exploration of ocean productivity, and factors that influence it on, scales relevant for fisheries worldwide. Workshops and pilot projects with the energy sector are exploring the use of km-scale global projections for these applications, and how their data can help assess the quality of the model output.
Network analyses of Hackathon participants documents how these new meeting formats — which encourage participation by researchers outside of nextGEMS — help foster collaboration and are nurturing a new pan- European analysis and development community. Use of video communication methods support outreach, heighten project visibility, and strengthen the communication skills of the nextGEMS research community.
The main limitation nextGEMS faces is access to computing resources. Despite the use of climate applications to justify Europe’s huge investment in HPC, nextGEMS is the only climate application to access these resources, and then only a very small fraction (less than a percent) of what is available.
Looking forward to the final third of nextGEMS, in addition to an increased emphasis on scientific exploitation and user engagement, new runs at higher (2.5 km horizontal grid spacings) are being performed. These runs, and their associated workflows, will be used to engage both an international scientific and user community through the globalization of the nextGEMS hack-a-thons with the culmination of the project.
Developing the models to run for climate applications identified many errors that had been hidden from previous generations of the models either by their exclusive use for short-timescales, or by the wide-spread practice of model tuning. By maintaining full control of the mass and energy budgets, and in particular physical processes influencing low clouds, both models could be shown to reproduce the annual budget of energy and temperature, and to run stably for periods of decades. Both models were run with a grid spacing of 10 km in the atmosphere and 5 km in the ocean for multi decadal (30 yr) projections based on a plausible emission scenario (SSP3-7.0) and the data has been published. The resolution now provides access to climate information for regions and on scales never before covered by future climate projections. Both models are being expanded to include new Earth-system related components. The fine scale of the models allows for a more physical coupling to a large variety of processes and establishes the storm-resolving modelling framework as a more physically compelling representation of the Earth system as a whole.
The models ability to exploit computational resources provided by EURO HPC has allowed the numerical experimentation required for their application to a variety of scientific questions. These include factors con- trolling the distribution of tropical precipitation, convective organization, low level stratiform clouds in the tropics, the energetic response to warming, the response to aerosol forcing, the effect of land-use changes on precipitation, and the influence of fine-scale heterogeneity on large-scale patterns of warming. Scientific findings include the important of air-sea energy exchange in low wind regimes for the distribution of tropical convection, a negative feedback between precipitation and soil moisture, and the importance of vertical mixing in regimes of stable stratification for controlling the distribution of cloud amount.
Storyline approaches have been guiding the use of output for applications with foci on renewable energy and food security. Coupled ocean-biogeochemical simulations enable a first exploration of ocean productivity, and factors that influence it on, scales relevant for fisheries worldwide. Workshops and pilot projects with the energy sector are exploring the use of km-scale global projections for these applications, and how their data can help assess the quality of the model output.
Network analyses of Hackathon participants documents how these new meeting formats — which encourage participation by researchers outside of nextGEMS — help foster collaboration and are nurturing a new pan- European analysis and development community. Use of video communication methods support outreach, heighten project visibility, and strengthen the communication skills of the nextGEMS research community.
The main limitation nextGEMS faces is access to computing resources. Despite the use of climate applications to justify Europe’s huge investment in HPC, nextGEMS is the only climate application to access these resources, and then only a very small fraction (less than a percent) of what is available.
Looking forward to the final third of nextGEMS, in addition to an increased emphasis on scientific exploitation and user engagement, new runs at higher (2.5 km horizontal grid spacings) are being performed. These runs, and their associated workflows, will be used to engage both an international scientific and user community through the globalization of the nextGEMS hack-a-thons with the culmination of the project.
When nextGEMS comes to an end, we expect to have:
• demonstrated an ability to stably run global coupled Earth system models for decades on horizontal grids between 2.5 km to 5.0 km.
• developed and prototyped the workflows for the use of such models in applications.
• documented robust improvements and lingering biases in the climatology that emerge at such scales and the degree of regional precision in their representation of regional climate.
• provided the first glimpse of how a representation of km-scale processes in the atmosphere, ocean, and in land surface, influence the climate system.
• illustrated the potential for further developing this new class of models to address the needs of users.
Achievements which go beyond the specific objectives of nextGEMS. Which can happen in part thanks to synergies with other projects that nextGEMS has helped incubate.
From a purely scientific perspective we expect to have
• provided the first tests of the hypothesis that storm-resolving models can well represent the mean tropical climate and its variability given a proper representation of atmospheric cloud radiative forcing using a coupled model.
• demonstrated the extent to which convective organization influences extreme rainfall as well as the atmospheric radiation budget.
• understood whether atmospheric moist convection couples fundamentally differently to the land surface and its implications for presumed critical thresholds in terrestrial ecosystems.
• provided the first quantification of climate sensitivity and aerosol effective radiative forcing from models that account for contributions from convective organization and convective scale processes.
Much of this work will be aided by a very large allocation of computer time on Europe’s largest pre-exascale machine (LUMI).
• demonstrated an ability to stably run global coupled Earth system models for decades on horizontal grids between 2.5 km to 5.0 km.
• developed and prototyped the workflows for the use of such models in applications.
• documented robust improvements and lingering biases in the climatology that emerge at such scales and the degree of regional precision in their representation of regional climate.
• provided the first glimpse of how a representation of km-scale processes in the atmosphere, ocean, and in land surface, influence the climate system.
• illustrated the potential for further developing this new class of models to address the needs of users.
Achievements which go beyond the specific objectives of nextGEMS. Which can happen in part thanks to synergies with other projects that nextGEMS has helped incubate.
From a purely scientific perspective we expect to have
• provided the first tests of the hypothesis that storm-resolving models can well represent the mean tropical climate and its variability given a proper representation of atmospheric cloud radiative forcing using a coupled model.
• demonstrated the extent to which convective organization influences extreme rainfall as well as the atmospheric radiation budget.
• understood whether atmospheric moist convection couples fundamentally differently to the land surface and its implications for presumed critical thresholds in terrestrial ecosystems.
• provided the first quantification of climate sensitivity and aerosol effective radiative forcing from models that account for contributions from convective organization and convective scale processes.
Much of this work will be aided by a very large allocation of computer time on Europe’s largest pre-exascale machine (LUMI).