Real-time state estimation of laboratory flows

Abstract

In this project, we use a real time computer model to simulate a differentially heated laboratory annulus. The laboratory annulus allows us to study chaotic flows typical of the atmosphere. Our objective is to bring the numerical model into close alignment with the laboratory system. Parameter estimation and data assimilation of real time model runs with the tank experiment can be used to improve numerical model fidelity, provided the model operates in real time and the model parameters are in acceptable regimes. We describe how to modify the default configuration of the MITgcm to tackle this new problem. We also run laboratory experiments. Using an iterative process, we update the model parameters (such as diffusion and viscosity), and observe that in at least some regimes, there is excellent agreement between observations and simulations. Much of this effort required the development of infrastructure, which is discussed in this document. Finally, we create and test a complete, real-time system, with data sent across the network from a parallel computer running the numerical model to the laboratory computer, where data assimilation takes place. We further modify the model to allow it to pause mid-run, and restart with the most recent state estimates of velocity and temperature.