[A-AS01] From Weather Predictability to Controllability

[AAS01-P05] Ensemble-Based Model Predictive Control for Meteorological Applications

*Kenta Kurosawa¹, Atsushi Okazaki¹ (1.Chiba University)

Model predictive control (MPC) is an optimization-based control framework, but its computational cost can be prohibitive for high-dimensional nonlinear systems due to the need for full model evaluations when minimizing the cost function. This study introduces ensemble model predictive control (EnMPC), a novel approach for nonlinear control that combines MPC and ensemble data assimilation. By solving the MPC cost function through ensemble approximations, EnMPC mitigates nonlinearity and uncertainty, improving computational efficiency over conventional MPC. To assess the potential of EnMPC for meteorological applications, we conducted numerical experiments using the Scalable Computing for Advanced Library and Environment (SCALE) model, focusing on the severe rainfall event of September 2015 in eastern Japan. Our experiments aimed to guide atmospheric conditions towards a state represented by an ensemble member exhibiting minimal precipitation impact, which we defined as the control objective. We successfully developed a prototype control framework for numerical weather prediction models, demonstrating its feasibility in influencing atmospheric conditions toward a desired state. In the presentation, we will introduce our findings from these experiments and discuss the insights into the practical implementation of the control system.