posted on 2020-03-10, 03:50authored byShona Mackie, Annika Seppala, Inga Smith
Understanding our climate and how it changes in future is a topic of increasingly urgent
research, with heightened levels of public and political support worldwide. Climate models,
however, are necessarily big. In theory, they represent all physical processes from the top of
the atmosphere to the bottom of the ocean, over land, water and sea ice, in 3-dimensional
grid cells with a resolution of 1 degree or finer. The interaction and evolution of these
processes is modelled with a temporal resolution of more than a single timestep per hour,
and typically we need to run for at least 100 years. Furthermore, uncertainties and internal
variability in the climate system mean that we run an ensemble rather than a single model
run. The structure of climate models means they can usually be parallelized to a point, but
they are not generally suitable for the distributed computing solutions that can be
implemented in other fields. As well as being expensive to run, climate models produce a lot
of data (PB scale). The idea is to capture the state of the whole world in a 3-dimensional
mesh with a temporal resolution fine enough to see how it changes, and a spatial resolution
fine enough to examine any physical process anywhere on Earth that might impact on
climate. For example, one model component (atmosphere, ocean etc) can be made of 1.2
million grid points. Saving just one parameter daily for 100 years = 44 billion data points. 30
parameters from 6 ensemble simulations amounts to 8 trillion data points, just from one
model component. These data have to be accessible so that we can do processing and
monitoring of climate model runs while they are underway, and need to be securely
archived in a way that makes them accessible for long term use, and shareable with
collaborators both present and future, here in Aotearoa and abroad. Running a climate
model is just the beginning of climate research, analysis of the data requires tools capable of
accessing and handling large data volumes that are generally stored on remote servers,
sometimes overseas, at a speed that makes interrogation and analysis practical.
Climate modelling is one of the most computationally hungry fields of research and New
Zealand has recently joined the list of the relatively few countries with the resources and
infrastructure to do it. Growing this field of research in New Zealand will need development
of resources and expertise to manage those resources. Events like eResearch 2020 are an important way for information to be shared with network architects and data managers to
ensure that the systems and infrastructure are in place to support the next generation of
climate researchers.
ABOUT THE AUTHOR(S)
Shona Mackie is a climate modeller at University of Otago, developing the New
Zealand Earth System Model to include new physics processes, and carrying out
senstivity studies using the current version of the model to better understand
uncertainties inherent in our climate projections.
Annika Seppälä is a senior lecturer at Otago University Physics department. Her
research uses computational simulations together with large space based Earth
observation datasets to investigate solar influence on the atmosphere and climate
from global to regional scales.
Inga Smith is a senior lecturer in the Department of Physics, University of Otago. Her
research interests are in sea ice physics and climate change, particularly the
influence of fresh water on sea ice formation.