Bayesian optimization of typhoon full-track simulation on the Northwestern Pacific segmented by QuadTree decomposition. Issue 208 (January 2021)

• Record Type: Journal Article
• Authors: Cui, Wei; Zhao, Lin; Cao, Shuyang; Ge, Yaojun
• Is Part Of: Journal of wind engineering and industrial aerodynamics. Issue 208(2021)
• Publication Date: 2021-01
• Subjects: Tropical cyclone -- Typhoon simulation -- Quadtree -- Bayesian optimization; Wind-pressure -- Periodicals; Buildings -- Aerodynamics -- Periodicals; Pression du vent -- Périodiques; Constructions -- Aérodynamique -- Périodiques; Buildings -- Aerodynamics; Wind-pressure; Periodicals
• DOI: 10.1016/j.jweia.2020.104428
• Languages: English
• ISSNs: 0167-6105

Abstract

Abstract: The tropical cyclone is one of the most destructive weather phenomena for several coastal countries, including China, the United States of America, India, Japan, and Australia. Accurate modeling of their trajectories is essential for public safety. Current research has provided several principle methods for simulating typhoon tracks and intensity development from genesis to landing and decaying. However, simulation performance still needs to be improved. This paper first presents a new ocean segmentation algorithm based on QuadTree to divide the analysis region in the Northwestern Pacific adaptively according to data sample density. The regression analysis area can be automatically adjusted to produce a significant enough data sample for fine-grained modeling. Second, it proposes a Bayesian optimization for parameter tuning. Because Month Carlo typhoon simulation results unavoidably incorporate uncertainties and need long computing time, Bayesian optimization is suitable for typhoon simulation parameter adjustments. With optimized parameters, the simulated typhoon activities demonstrate better agreement with historical records. Highlights: QuadTree structure is used to divide the ocean adaptively according to data density. Ocean is divided into nonuniform grids for typhoon parameter regression analysis. Robust linear regression is employed to improve simulation parameters reliability. Bayesian optimization is performed to tune simulation parameters to improve accuracy.