Take a look at this amazing simulation of how carbon dioxide and carbon monoxide gases change in volume and move through the atmosphere over time.
The swirling red/orange/yellow colors in the northern hemisphere represent carbon dioxide while the white/gray/black colors in the southern hemisphere represent carbon monoxide. Over the course of the simulation you can see how atmospheric conditions affect the movement of greenhouse gases away from their source to other regions on the planet and how the volume of these gases in the atmosphere changes with the seasons. Note the extraordinary reduction in carbon dioxide in the norther hemisphere as summer plant growth sucks the gas out of the atmosphere.
The computer model that produced this simulation was developed at NASA’s Goddard Space Flight Center in Greenbelt Maryland. The model combines real-world data on man-made and natural greenhouse gas emissions with atmospheric conditions and then simulates changes over time. The simulation in the video runs from May 2005 to June 2007.
As a personal aside, my normal daily cycling route takes me around part of the perimeter of NASA’s Goddard Center in Greenbelt. Until a few years ago when they reworked the surrounding road grid for security purposes, my route took me right through Goddard.
The simulation model operates at an extraordinary resolution. Typical global climate models are based on a horizontal grid of boxes that are 50 kilometers (a bit more than 31 miles) long. The NASA simulation makes use of a grid of 7 kilometers (a bit more than 4.3 miles) boxes. In other words, a 50k box that is represented by 1 data point in a typical global climate model is represented by approximately 49 separate data points in the NASA model. The simulation you see in the video produced almost 4 petabytes of data (1 petabyte = 1 million X 1 billion bytes = 1,000,000,000,000,000 bytes). Completing the simulation took 75 days of runtime.
For an illustration of how far Moore’s Law has taken us over the past 30 years, compare the NASA simulation model with the human memory models I was building in graduate school in the 1980s. I was fortunate to be working in a lab that had, at the time, state-of-the-art computer resources in cognitive psychology – a pair of DEC PDP 11-34 computers. People would regularly stop by our lab to marvel at our equipment. I was writing code (in Fortran) for these beauties that ran computer-controlled experiments with human participants, and also ran simulations based on a variety of human memory models and early neural network models. The workspace we had available to us in the PDP 11-34s for the models, the simulation and experimental runs, the human and simulation data, and the data analysis programs was 64 kilobytes (approximately 64,000 bytes). We thought it was luxurious.