14:00 〜 14:15
[AAS01-02] Exploring weather control technology to steer the atmosphere towards favorable directions based on ensemble data assimilation
キーワード：気象制御、データ同化、モデル予測制御、豪雨
Japan's Moonshot program, initiated by the Japan Science and Technology Agency, aims to mitigate weather-induced disasters by developing weather control technology. Our project aims at exploring methods for generating upstream maritime heavy rains to govern intense-rain-induced disasters over land. The atmosphere that brings heavy rain is conditionally unstable, and therefore, artificial interventions could induce convections intentionally. In our project, we are developing methods for optimizing places and timings of the interventions for organizing the convections.
Specifically, we focus on model predictive control (MPC) to optimize interventions so that the atmosphere follows a targeted trajectory. This optimization problem can be viewed as data assimilation. In numerical weather prediction (NWP), data assimilation synchronizes simulation worlds with the real atmosphere. In our weather control project, we first identify a desirable trajectory with reduced disaster risk, then the MPC synchronizes the real atmosphere with this trajectory through optimized interventions.
Our initial investigations involve analyzing latent-space trajectories to find a separatrix that differentiates disaster and non-disaster scenarios. Using operational ensemble weather prediction data provided by Japan Meteorological Agency, our data-driven approach captures latent-space trajectories, revealing two distinct regimes that may be controllable with small manipulations. Additionally, we have begun coupling the model predictive control and data assimilation to find effective interventions that lead the atmosphere towards prescribed directions. This presentation will provide an overview of our research, and the most recent advancements until the time of the conference.
Specifically, we focus on model predictive control (MPC) to optimize interventions so that the atmosphere follows a targeted trajectory. This optimization problem can be viewed as data assimilation. In numerical weather prediction (NWP), data assimilation synchronizes simulation worlds with the real atmosphere. In our weather control project, we first identify a desirable trajectory with reduced disaster risk, then the MPC synchronizes the real atmosphere with this trajectory through optimized interventions.
Our initial investigations involve analyzing latent-space trajectories to find a separatrix that differentiates disaster and non-disaster scenarios. Using operational ensemble weather prediction data provided by Japan Meteorological Agency, our data-driven approach captures latent-space trajectories, revealing two distinct regimes that may be controllable with small manipulations. Additionally, we have begun coupling the model predictive control and data assimilation to find effective interventions that lead the atmosphere towards prescribed directions. This presentation will provide an overview of our research, and the most recent advancements until the time of the conference.