非线性驱动LGCP模型的台风路径演化机制分析

Analysis of Typhoon Track Evolution Mechanisms Using Nonlinearly-Driven LGCP Models

  • 摘要: 基于2020-2024年中国台风路径点数据,结合温度场、湿度场、风场等环境协变量,构建了广义加性模型(Generalized Additive Model,GAM)框架下的Log Gaussian Cox Processes(LGCP)耦合模型,为台风路径预测和演化机制分析提供了新的思路和方法。主要研究内容包括:(1)将GAM的非线性建模能力引入LGCP框架,有效解决了传统GLM模型对台风-环境复杂关系的刻画不足问题;(2)提出了基于张量积平滑的风场因子交互效应建模方法,量化了垂直风切变与低层风速的协同贡献(25.12%)。研究结果表明:台风路径点空间相关性在20°范围内,GAM模型较GLM提升解释率12.45%,其中海温(26.73%)与湿度(25.29%)的非线性效应及风场协同作用(25.12%)构成关键驱动。该研究不仅为台风灾害风险评估提供了更可靠的空间统计方法,其提出的“环境因子-空间过程”耦合建模框架,揭示了台风路径对环境的非线性效应,对完善热带气旋生成理论具有重要科学价值,也为极端天气事件预测、沿海地区台风灾害预警系统优化等提供可行有效的模型方法。

     

    Abstract: Objectives: Typhoon track prediction is crucial for disaster prevention and mitigation. This study aims to develop a statistical framework that integrates environmental covariates to enhance the accuracy of typhoon track forecasting and improve the understanding of underlying evolutionary mechanisms. Methods: A coupled Log Gaussian Cox Process (LGCP) model within a Generalized Additive Model (GAM) framework was established using 2020-2024 typhoon track data from China along with environmental covariates including temperature, humidity, and wind fields. Key methodological innovations include: (1) integrating GAM's nonlinear modeling capability into the LGCP framework to overcome the limitations of traditional GLM in capturing complex typhoon-environment relationships; (2) developing a tensor product smoothing approach to model interaction effects among wind field factors, enabling quantitative assessment of synergistic effects between vertical wind shear and low-level wind speed. Results: The analysis revealed a significant spatial correlation range of 15-25 degrees in longitude and latitude for typhoon track points. The GAM model demonstrated a 12.45% improvement in explanatory power compared to GLM, with sea surface temperature (26.73%), humidity (25.29%), and wind field interactions (25.12%) identified as primary driving factors. The synergistic effect between vertical wind shear and low-level wind speed was quantitatively evaluated at 25.12%. Conclusions: This research provides an advanced spatial statistical tool for typhoon risk assessment and disaster management. The proposed "environmental factor-spatial process" coupled modeling framework offers a universal methodology for extreme weather prediction and climate adaptation planning. Theoretically, it reveals nonlinear environmental effects on typhoon tracks, cficantly to tropical cyclone genesis theory. Practically, it supports the optimization of early warning systems for coastal typhoon disasters and facilitates the implementation of national strategies for building climate-resilient cities.

     

/

返回文章
返回