The Crosscurrents event was a marine science and poetry event held at Plymouth Marine Laboratory (PML) in March 2017. At the event, five scientists were paired up with five local poets. The scientists were tasked with presenting their research and the poet shared one or more poems inspired by the science. On the night of the event, there was a fantastic turnout, filling the PML lecture theatre.
Each pair of scientists and poets presented their work together. I represented the UK Earth System Model and I was paired with the poet Ben Smith from the English department of the University of Plymouth. Rather than discussing the results of previous generation models or the study of climate change, we instead decided to use the Crosscurrents event to discuss lesser known aspects of Earth System models. We focused our presentation on the complexity required of the UKESM, the process of model development, and our current progress on the ocean spin up; topics which are rarely discussed with non-experts. We tried to capture these lesser discussed elements of climate science because we felt that the public are often only presented with the results of the work, and never hear about the methodology. In this article, we present two aspects of model development and their associated poems: the iterative process of model development and spinning up the simulated ocean’s carbon cycle.
While building the model, it’s crucial to know how new developments change the model results. The typical work process of a modeller is: run the model, then check the model results against recorded data or previous model runs. If the model does not behave similarly enough to recorded data or is noticeably worse than the previous model runs, then there is a problem and we stop the run. After identifying the problem, we make a change to the model’s code, then restart the run from the beginning. This iterative process means that we delete plenty of failed simulations. This idea inspired Ben to pen the following poem, Alternate Histories:
What I won’t think about are the glitches in the system.
The one where, after four hundred years of smooth running,
the ocean disgorged its carbon skywards
like iron filings drawn up by a magnet.
I won’t dwell on it.
I won’t dwell on the model where a sudden storm
deposited a cloud of sediment
causing a plankton bloom
that fused all the world’s water into a stiff, organic paste.
I won’t think of that paste,
its texture or smell
and I won’t think of that other model
where, after the smallest change
in the initial calculations,
all marine life evaporated
leaving the sea bright as a polished lens.
Instead, I’ll look out of my dusty window
at the blue-tit hanging on the back of a fern,
the plum tree teetering on the edge of spring,
and try to find other words for
try not to think of the hard drives
filled with terabytes of failed worlds
that never even made it to now.
Another aspect of the day to day life of an Earth system model developer is spinning up the model. We can’t simply start the model one day and look at the output data of the next day. As you know, today’s weather depends on yesterday’s weather. If it is raining here in Plymouth, the clouds that drop the rain were likely brought here from the sea to the west. It was probably bright and sunny somewhere over the ocean sometime last week. The sunlight warmed the sea surface, causing some of the water to evaporate. The water vapour was condensed into clouds, carried here by winds, and then rain fell.
If today’s weather depends on yesterday’s weather, then yesterday’s weather depends on the day before and so on. Effectively, the climate has a memory of what has happened in the past. This means that in order to study climate change, we must gradually introduce climate change gases to a natural-like simulated environment, as similarly as possible to how it actually happened between the start of the industrial revolution and modern day.
In addition, if we start the model from a standstill, it can be pretty chaotic at the beginning. Imagine this: the atmosphere is entirely completely still, no wind, the sea is flat, the Earth isn’t turning, there’s no light. Suddenly, the sun switches on, the Earth starts spinning, the wind starts blowing and the ocean starts sloshing around. It’s not a very natural scenario, but this is what happens if we start a model from standstill. It may take some time to settle down to a sensible steady state behaviour that we can use as a starting point for the industrial revolution.
Furthermore, in the ocean, cold dense water sinks and warm buoyant water floats. Once some cold dense water sinks, it can take thousands of years before it comes back up to the surface again. This means that the current state of the ocean can be influenced by events from thousands of years ago!
So, with all that in mind, how long do you need to run the model before it “forgets” the initial conditions? The answer is that we need to spin up for many thousands of years to reach a steady state, before we are ready to subject the simulation to a gradually changing climate. Ben Smith presented what is likely to be the first poem ever on model spin up, Spinning Up:
The way a spinning top achieves stillness
settling in motion
parallel to the table’s surface
so the model is set running
and must hold itself steady
in relation to the world.
This is what is needed for prediction:
numbers held in balance
with the data
of the present moment.
But, how to know precisely
when we’ve reached
this point of mirroring,
when each flicker of movement
is nothing more
than our own eyes flickering?
Is it now
At the event, Ben also presented two other poems, and the combined works of the Crosscurrents project, including contributions from the other scientists and poets will be published at a later date. This was a unique, fun, thought provoking and engaging event, and I would encourage other scientists to participate in collaborative art-science outreach projects. For more information about the Crosscurrents Project, please contact Dr. Ben Smith: email@example.com