Controlling chaos in El Niño

Many weather and climate phenomena are chaotic in nature; indeed for many people this is the canonical example of a chaotic system. However, because of this, it is at least theoretically possible to have significant influence over these systems with extremely small control inputs. This potential is explored using the Cane-Zebiak 33 000-state model of the El-Niño/Southern Oscillation (ENSO). The model dynamics are nonlinear and chaotic, and the optimal control input can be found through iteration using the adjoint simulation. The performance of this optimal control (which implicitly assumes perfect model and state information) is compared with a simple SISO linear feedback. Significant reductions in ENSO amplitude are (theoretically) possible with very small control inputs, illustrating that it is possible to have significant influence over large-scale climatic phenomena without correspondingly large control effort.