High performance computing (HPC) will drive a sharp increase in
our ability to more accurately model and predict climate change
effects, as well as other complex systems. One of the biggest
challenges about understanding the effects of climate change and
working to mitigate them is that the models used to predict the
measures of climate change (temperature, ocean currents,
atmospheric circulation, etc) are quite inaccurate. Smith et al noted
in a 2020 paper that "modelling and internal variability uncertainties
are each large enough to allow opposite projections of European
winters" particularly at the decadal timescale. UN projections, which
are the basis of many countries' climate policies, predict change on
the 100-year timescale, which is, frankly, too long to be useful, and
on a global dimension, which is too general to provide explanatory
power for regional effects of climate change, which can vary
substantially by latitude, geography, etc.
High performance computing will make it possible to do a couple of
things that should make climate modeling exponentially more
accurate and more available. First, the amount of data, and the
myriad of types of data, that are needed for complex models is
enormous, and the task of data cleansing and preparation laborious.
This is, for most organizations, prohibitively expensive (Vogels,
2020). Second, the amount of compute power needed to run a
complex model is also massive and difficult & expensive to procure
(Vogels, 2020). In recent years, this technology (e.g. distributed high
performance computing, and the ability for multiple models to feed
each other in real time) has migrated out of the federal/academic
research space and into the commercial sphere, where it is
becoming more of a commodity. e The ability to use these kinds of
tools will make it easier to get better climate predictions that can
then be used to refine the policy options available and hopefully
drive improved decision making.
Smith, D.M., Scaife, A.A., Ead, R., et. al. (2020 July 29). North
Atlantic climate far more predictable than models imply. Nature.
583, 796-800. https://doi.org/10.1038/s41586-020-2525-0
Vogels, W. (2020 November 19). Understanding climate change
using high performance computing and machine learning. All Things
Distributed (Blog).
