Progress Report
Last updated:
Artificial generation of upstream maritime heavy rains to govern intense-rain-induced disasters over land (AMAGOI)[1] Control Theory
Progress until FY2025
1. Outline of the project
There are three hurdles in manipulating weather. First, we need to find effective interventions, as an ill-planned intervention may not produce the desired outcome. Secondly, interventions must be feasible; weather, being large-scale, may require enormous amounts of energy to control, while there are also constraints on the energy we humans can expend. Finally, the intervention must be computable; interventions must be computed in the short time, before heavy rainfall damage is predicted. Although the control theory can be essential to overcome these three hurdles, the current control theory has not dealt with such large-scale phenomena.
To clear these hurdles, two research tasks have been established. Subject 1-1, Development of Feedback Control Theory, extends the concept of feedback control and develops intervention-design methods scalable for large-scale weather fields. Subject 1-2, Development of Data-driven Control Theory, develops weather control methods that use AI techniques, ensemble prediction, and related data technologies to identify effective intervention candidates efficiently. Through these subjects, we are moving from idealized atmospheric experiments toward more practical weather fields, while aiming to demonstrate the feasibility of mitigating heavy rainfall over land through offshore rainfall generation on a computational basis.
2. Outcome so far
[Model Selection (Sub.1-1)]
Effective weather control research requires models both tractable and representative of real atmospheric phenomena. We adopted real atmospheric experiments within the SCALE Regional Model, which offers both realism and usability and, hence, is well-suited to advance our study.
[Evaluation of Existing Control Methods (Sub. 1-2)]
We evaluated a conventional weather intervention method [Miyoshi and Sun, 2022] and showed, through comparison with Model Predictive Control, that its reliance on the prediction window leads to inherent limitations.
[Theory for Overcoming Large-Scale Challenges (Sub. 1-1)]
Meteorological phenomena are large-scale, making the direct application of traditional control methods computationally challenging. We have developed a fast intervention calculation method based on convex optimization. By preparing perturbation analysis in advance, interventions can be computed in a short time even in Ideal Atmospheric Experiments with thousands of state dimensions (Fig.1). We are also examining feedback-type computation in which the degrees of freedom of intervention are restricted to reduce sensitivity-analysis costs, while interventions are updated according to the atmospheric state at each time step.
[Intervention Search via Expensive Optimization (Sub. 1-2)]
To reduce rainfall through weather control, many simulations are needed; however, conventional methods struggle with high computational costs. We applied Bayesian Optimization-based Expensive Optimization to efficiently identify effective interventions from limited trials, considering both one-shot and sequential intervention settings. The reason for the effectiveness of Bayesian Optimization lies in its ability to select promising interventions based on the results of previous simulations while also exploring new possibilities. We are also studying approaches that use physical information to narrow the search space and update interventions based on future predictions. These results suggest a pathway toward closed-loop weather control under limited computational resources (Fig.2).
3. Future plans
We aim to extend the developed weather control methods to large-scale, more realistic simulations for mitigating heavy rainfall over land under limited computational resources. We will compare the developed approaches in terms of robustness, scalability, and efficiency, and advance real-time Model Predictive Control based on forecast data to demonstrate feasible rainfall mitigation.