The world's climate scientists are charged with a difficult task:
to create a crystal ball with which to skry a future that promises to be hotter than today.
But exactly how much hotter depends on innumerable factors, both natural and human.
Creating the crystal ball is thus a two-stage process. First, you have to build a simulacrum of how Earth's climate works.
Then, you try to perturb this simulacrum with plausible future human actions, to see what picture appears.
Modern magic being what it is, the crystal balls are actually supercomputers running programs with 1m or more lines of code.
These programs are models that divide the planet's atmosphere, ocean and land surface into grids of cells—many millions of them.
Land cells are flat. Atmosphere and ocean cells are three-dimensional
and are stacked in columns to account for the effects of altitude and depth.
A model calculates what is going on, physically and chemically, inside each cell,
and how this will affect that cell's neighbours, both sideways and, if appropriate, above and below.
Then it does it again. And again. And again. That is a complicated process.
A model's code has to represent everything from the laws of thermodynamics to the intricacies of how air molecules interact with one another.
Running it means performing quadrillions of mathematical operations a second—hence the need for supercomputers.
And using it to make predictions means doing this thousands of times,
with slightly different inputs on each run, to get a sense of which outcomes are likely,
which unlikely but possible, and which implausible in the extreme.
Even so, such models are crude. Millions of grid cells might sound a lot,
but it means that an individual cell's area, seen from above, is about 10,000 square kilometres,
while an air or ocean cell may have a volume of as much as 100,000 cubic kilometers.
Treating these enormous areas and volumes as points misses much detail.
Clouds, for instance, present a particular challenge to modellers.
Depending on how they form and where, they can either warm or cool the climate.
But a cloud is far smaller than even the smallest grid-cells, so its individual effect cannot be captured.
The same is true of regional effects caused by things like topographic features or islands.