Climate Models: Mirages for Virtual Climatic Realities

“All climate models are wrong, but some are useful” Statistician George Box (1919-2013).

I am a climatologist. I use a climate model. This is my laboratory where I perform all my experiments.

This doesn’t mean I bounce a jelly Mini-Earth in the lab, it means I describe Earth’s climate system through a climate model made of several computer programs to create a virtual Earth. Inside this virtual reality, I aim to understand the changing processes on the Earth’s climate system through. Processes like clearing farming land to build cities for a larger population in Brazil or intense deforestation in Indonesia. Others including more/less industrial emissions into the atmosphere/ocean creating fog and pollution in China or India or the extent of Antarctic/Artic sea ice & sea-level rise in Tuvalu. This is the only tool I have to better understand our system; because I cannot conduct large scale experiments on the vast atmosphere itself; nor am I a time traveler (to see past/future changes). Is this concept so different from playing computer games like MY2050, CO2FX or SimCity?

The climate model (collection of math equations to describe the different (physical, chemical, geological) processes that drive the Earth’s climate) is run on a supercomputer; demanding huge amounts of expensive computational resources. The science of climate modelling stems from laws of physics (like Newton & and laws of thermodynamics). Each equation contains many variables like temperature, rainfall, sea-level rise, and when we combine all these equations, through individual & collective interactions, we see how the climate evolves (atmosphere, ocean, land, sea ice & the sun) from the North Pole to the Kiribati Islands (equator) and the South Pole. This is a complex & daunting task. 

Fig 1: Running computer codes and checking experimental outputs via remotely connected to the supercomputer 

Information of processes are resolved by 3 dimensional grid boxes (cubes) called spatial model resolution (pixels/ grid points) with larger grids at the equator and smaller at the poles (which can be made of several kilometers). A 2-degree spatial resolution specifies horizontal grid box of ~ 210km but even at 1-degree horizontal resolution (111km) it is still too coarse (Fig2) to capture finer processes like ocean eddies/ topography (mountains) or evaporation over the Easter islands. Choices for resolution depends on available computing power and the time needed to run the model experiment. In the current golden age of high resolution (smaller grids), the more complicated & higher resolution a model, the slower it is on the supercomputer. Choices are made in every aspect of building, running and analyzing climate models for efficiency. But as climate models (regional to global scales) span longer periods (years, decades, or millennia) than weather models (local scales), they cannot include as much detail. Most scientists agree that high-resolution models were far better at reproducing observational data (e.g. from satellite or gauges) & intense storms, cyclones & hurricanes (fine scale climate events). This is not always true. But to have a perfect simulation of reality, climatologists need understanding of each process (every cause and effect). A question to ponder, won’t there always be processes we don’t fully understand but think are important? Processes like clouds which are on smaller resolutions then model. When we achieve the finest resolution, would these climate models be simulators of reality then? 

Modern day climate model descendants from the first pioneering model (1960s) at NOAA Geophysical Fluid Dynamics Laboratory in Princeton, New Jersey have improved vastly. These models agree/disagree with each other due to choices made when incorporating elements in building each model, such as the treatment of clouds, aerosols and the carbon system. This gives a different virtual reality each time. The Community Earth System Model (CESM1.2.2) is one of these ~30 climate models. The CESM is unique as it is developed by a broad community of scientists & freely available to researchers worldwide. It encompasses a much improved holistic science on major components of the climate system. This gives a better representation of the Earth’s climate system. Using the CESM, I hope to learn more about large scale ocean-atmosphere teleconnection patterns such as the North Atlantic Oscillation and the El Nino Southern Oscillation. These affect sea surface temperatures as well as atmospheric conditions such as change in pressure and winds which are responsible for rainfall. This can lead me to understand the past extreme events or predicting future years of potential impacts, like a high probability of drought (warm and dry) for the Caspian. Or several years of cold and wet conditions (flooding) with catastrophic implications for local people, agriculture & economy through time.

How do climate modellers predict the future?

We do not.

Predicting the future is impossible. Instead we make climatic projections for a set of different possible futures (creating virtual realities) based on physics and predications of future CO2 and other emissions. When modelling a century for these virtual realities, even the 5th Intergovernmental panel on climate change (IPCC) offers “what if” projections of future climate scenarios that relate to certain emissions scenarios. Inside this virtual reality, I can tell how hot the next 5-10 summers will be, but not how hot a weekday in that one summer. There are limitations; even inside this virtual reality.

We make assumptions on how the earth system works. These assumptions are simplified, because the climate is complex and computing power limited. The truth is complex and models are merely an approximation of the truth. With a little bit of patience, time, more data to work with and more powerful computers, climate models will improve through advanced scientific understanding of Earthly processes. But skeptics continue discrediting climate models, basing it on crudity & simplification reflected in reality. Hence, rather than viewing models as the literal truth, we should view it as alternate realty which is something useful. As featured in Before the Flood documentary, Piers Sellers, the British born astronaut, acclaims that “as the science community, we have not done the best job, frankly, of communicating this threat to the public. When you go up there and see it with your own eye, you see how thin the world’s atmosphere a tiny little onion skin around the Earth”.

Climate modellers, as humans, have limits to their scientific understanding and computing power, hence “all models are wrong” because they are a simplification of reality “… but some are useful.” These simplifications are the only tools to explain, predict & understand the climatic process on Earth. Does a climate model have to be perfect to be useful? The hard part is assessing whether a model is a good tool for the job at hand. So the question for my next blog is:

How do we assess the usefulness of a climate model (CESM); for the Caspian in particular? We show ways to quantify climate modelling uncertainties and its impacts, past and future.

By: Sri Nandini

  • PRIDE early stage researcher
  • Doctoral Candidate, 
  • MARUM – Center for Marine Environmental Sciences
  • Faculty of Geosciences, University of Bremen, Germany.