基于Vickery模型的气候变暖台风路径模拟

Typhoon Track Simulation under Global Warming Scenarios Based on Vickery Model

摘要:

为探讨不同气候变暖模式下对未来台风路径的潜在影响，基于 CCCma 提供的在三种气候变暖假设下未来百年间海平面温度的预测值，完成了西北太平洋的全路径台风模拟。每年台风发生的次数按照历史观测值的 PDF （时不变的负二项分布）随机模拟。台风模拟的路径、强度模型沿用 Vickery 模型，对于台风梯度风场的模型，在 Georgious 梯度风风场模型的基础上进行改进，基于神经网络模型（ANN）重新拟合了最大风速半径的关系式，模型输入有台风中心压差、纬度和海平面温度，可以将气候变化的影响也纳入到时变台风年极值风速 PDF 中。揭示了气候变化导致的海平面温度升高对台风登陆频率、台风移动速度、台风移动方向和台风中心压差的可能影响。结果表明，气候变暖对年极值风速 PDF 的影响具有地域性，大部分地区年极值风速 PDF 均值会变大，离散性也会更大，少部分地区的长期趋势未受气候变暖影响。采用 Gumbel 分布对年极值风速 PDF 进行拟合，Gumbel 分布的位置参数和形状参数具有时变特性并通过正态分布量化线性拟合的误差，并称之为估计不确定性。

Abstract:

To investigate the potential effect of different global warming scenarios on future typhoon track, this study utilized the predicted values of sea surface temperature (SST) over the next century under three global warming scenarios provided by CCCma to conduct full-track typhoon simulations in the Northwest Pacific. The annual frequency of typhoons was randomly simulated according to the probability density function (PDF) of historical observations, which followed a time-invariant negative binomial distribution. Typhoon tracks and intensities were simulated using the Vickery model. For the gradient wind field model of typhoons, improvements were made based on the Georgious gradient wind field model. An artificial neural network (ANN) was used to refit the relation for the radius of maximum wind speed, with typhoon central pressure deficit, latitude, and SST as inputs. This enabled the integration of climate change effects into the time-varying PDF of annual extreme wind speeds. It revealed the potential effects of SST increases caused by climate change on the frequency of typhoon landfalls, typhoon movement speed, typhoon movement direction, and typhoon central pressure deficit. The results showed that the effect of global warming on annual extreme wind speed PDF exhibited regional variability. In most regions, the mean and dispersion of the annual extreme wind speed PDF tend to increase, while in a few regions, the long-term trend remained unaffected by global warming. The Gumbel distribution was employed to fit the annual extreme wind speed PDF. Both the location and shape parameters of the Gumbel distribution exhibited time-varying characteristics. The errors in linear fitting were quantified using a normal distribution and were referred to as estimation uncertainty.