Climate/weather models are essential tools for predicting and understanding the Earth’s climate/weather. These models divide the globe into pre-defined numerous grid points or boxes to solve complex physical equations using numerical approximations. The concept of mathematical weather prediction was first introduced by Lewis Fry Richardson in 1922 [1], long before fast computing systems were available. Richardson envisioned a pool of humans placed at each grid point, simulating the role of modern computer CPU cores/nodes. This imaginative representation laid the foundation for the development of weather and climate models.
In 1969, the first atmosphere-ocean prototype model was developed and used by Manabe and Bryan. This early model, run on a UNIVAX 1108 computer, took approximately three and a half months of real-world time to simulate just one year of simulated climate/weather conditions[2]. The globe was divided into grid points spaced 500 kilometers apart, with 9 vertical height levels. While this model provided valuable approximations of weather and climate patterns at each 500-kilometer grid box, it inherently missed crucial information occurring within the spaces between these grid points. For instance, a tornado or a dust storm could occur in an area not accounted for by the model, leading to inaccuracies in the simulation.
To address this limitation, scientists introduced sub-grid parameterizations in the models. These are simple numerical approximations used to account for missing processes that occur between the grid points. While these parameterizations help improve the accuracy of climate predictions, they are not as precise as directly resolving the phenomena in question. Some of the examples of sub-grid scale parameterizations include convection (rise of warm air), turbulence, cloud processes etc.
With advancements in supercomputing facilities round the world, researchers have made significant strides in improving model resolution. By dividing the globe into smaller grid points, such as 5 kilometers or even 1 kilometer, high-resolution models[3]can capture more detailed climate processes and reduce the need for sub-grid parameterizations. These improvements hopefully allow for more reliable climate predictions.
In our lab, we have developed simulations using different resolutions by dividing the globe entirely covered with water (known as aqua-planet) into 5-kilometer and 40-kilometer grid points. By comparing these simulations, one can easily notice the enhanced detail and realism of low level clouds in the 5-kilometer run!
