Improving Tropical Cyclone Intensity Forecast Through Uncertainty Estimation and Long-Term Priors
Zihao Xu, Renlong Hang, Jiuwei Zhao, Gaozhen Nie, Qingshan Liu
Nanjing University of Information Science and Technology China Meteorological Administration Nanjing University of Posts and Telecommunications
阅读操作
确认中在文库中上传 PDF 后可生成中文音频讲解。
摘要与影响
In recent years, deep learning (DL) has made considerable progress in tropical cyclone (TC) intensity forecast. However, the existing DL models are constrained by their deterministic nature and a lack of modeling for long-term evolution patterns, thus precluding uncertainty estimation and limiting their generalization on complicated intensity evolution. To overcome these deficiencies, we propose a DL model by incorporating uncertainty estimation and long-term priors. Through learning distributional representations to jointly estimate data and output uncertainty and encoding historical patterns into graph convolution layers as a prior, our model enhances its 24-h forecasting performance, showing promising capability in handling complex intensity fluctuations. To evaluate the performance of our proposed model, we collect TC data from 2000 to 2022 and compare it with several state-of-the-art DL and operational models. Experimental results show that our model can reduce the forecasting errors by 14.35% and 9.91% in comparison with the best-performing operational model China Meteorological Administration (CMA) and the leading DL baseline (VQLTI), respectively. The code of our proposed model will be released athttps://github.com/captain-slow-tc/tcip-lpue
逐年被引趋势
暂无年度引用数据
关键指标
同类平均 = 1
同领域 · 同年份 · 同类型
Google Scholar 与 OpenAlex 的被引统计范围不同,数值存在差异属正常。
AI 辅助阅读
依据:摘要
可就本文提问;依据不足时会说明。
学术脉络
学科主题
物理Tropical and Extratropical Cyclones Research
Meteorological Phenomena and Simulations · Traffic Prediction and Management Techniques