一種基于深度學習模型的雷達回波臨近外推預報方法

doi:10.12306/2025jms.0012

首頁 > 過刊瀏覽>2025年第45卷第4期 >549-559. DOI:10.12306/2025jms.0012

一種基于深度學習模型的雷達回波臨近外推預報方法
DOI:
                        10.12306/2025jms.0012
                    
CSTR:
                        
                    
作者:
                        
                        
                    
作者單位:
作者簡介:
通訊作者:
中圖分類號:TP456
基金項目:中國氣象局創(chuàng)新發(fā)展專項(CXFZ2023J008)；山東省自然科學基金面上資助項目(ZR2021MD121;ZR2022MD072;ZR2022MD088)；山東省氣象局榜單類專項(2023SDBD01)；海河流域氣象科技創(chuàng)新資助項目(HHXM202404)；環(huán)渤海區(qū)域海洋氣象科技博同創(chuàng)新項目(QYXM202301)

A deep learning-based extrapolation method for radar echo nowcasting

Author:

Affiliation:

Fund Project:

摘要

圖/表

訪問統(tǒng)計

參考文獻

相似文獻

引證文獻

資源附件

文章評論

摘要:

針對氣象雷達回波深度學習外推方法中存在的回波邊緣扭曲、模糊彌散失真及真實性衰退的問題，在卷積長短期記憶網(wǎng)絡(Convolutional Long Short-Term Memory，Conv-LSTM)中引入殘差模塊、生成器與判別器，構(gòu)建生成對抗—殘差卷積長短期記憶網(wǎng)絡(Generative Adversarial-Residual Convolutional Long Short-Term Memory Network，GAN-rcLSTM)深度學習模型，同時針對不同強度回波賦予不同權重得到自定義的加權損失函數(shù)并融入GAN-rcLSTM得到自定義損失函數(shù)—生成對抗—殘差卷積長短期記憶網(wǎng)絡(Weighted Loss Function-Generative Adversarial-Residual Convolutional Long Short-Term Memory Network，Wloss-GAN-rcLSTM)?；?021—2022年山東省及周邊地區(qū)歷史雷達回波數(shù)據(jù)集對Wloss-GAN-rcLSTM模型進行回波外推訓練與測試，建立預報時長為0～2 h、逐6 min更新的時空型深度學習雷達回波外推模型。檢驗評估表明，Wloss-GAN-rcLSTM雷達回波外推模型較光流法、預測遞歸神經(jīng)網(wǎng)絡(Predictive Recurrent Neural Network，PredRNN)及GAN-rcLSTM方法在強降水關注的45 dBZ閾值下其臨界成功指數(shù)(Critical Success Index，CSI)分別提升了0.12、0.07與0.02，清晰度指標結(jié)構(gòu)相似性指數(shù)(Structural Similarity Index，SSIM)分別提升了0.009、0.042與0.11，且個例檢驗顯示W(wǎng)loss-GAN-rcLSTM最優(yōu)適用于颮線等中尺度天氣過程預報。

Abstract:

To address issues such as edge distortion， blurring， and loss of realism in radar echo extrapolation using deep learning methods， a residual module， generator， and discriminator were introduced into the Convolutional Long Short-Term Memory (Conv-LSTM) framework to construct a Generative Adversarial-Residual Convolutional Long Short-Term Memory Network (GAN-rcLSTM) deep learning model. Additionally， a customized weighted loss function， which assigns different weights to radar echoes of varying intensities， was designed and integrated into GAN-rcLSTM， resulting in the Weighted Loss Function-based Generative Adversarial Residual Convolutional Long Short-Term Memory Network (Wloss-GAN-rcLSTM) model. Using a historical radar echo dataset from Shandong Province and its surrounding areas (2021-2022)， the Wloss-GAN-rcLSTM model was trained and tested for radar echo extrapolation. A spatiotemporal deep learning radar echo extrapolation model capable of 0-2 hour forecasting with 6-minute updates was established. Evaluation results indicate that， at the Critical Success Index (CSI) threshold of 45 dBZ， which is of particular interest for heavy precipitation， the Wloss-GAN-rcLSTM model outperforms the optical flow method， Predictive Recurrent Neural Network (PredRNN)， and GAN-rcLSTM by 0.12， 0.07， and 0.02， respectively. For the clarity metric， structural similarity index (SSIM)， improvements of 0.009， 0.042， and 0.11 units were observed， respectively. Case studies further demonstrate that Wloss-GAN-rcLSTM is effectively suited for forecasting mesoscale weather processes， such as squall-line systems.

參考文獻

相似文獻

引證文獻

引用本文

魏海文,郭俊建,周成,王靚,張登旭.一種基于深度學習模型的雷達回波臨近外推預報方法.氣象科學,2025,45(4):549-559 WEI Haiwen, GUO Junjian, ZHOU Cheng, WANG Liang, ZHANG Dengxu. A deep learning-based extrapolation method for radar echo nowcasting. Journal of the Meteorological Sciences,2025,45(4):549-559

復制

文章指標

點擊次數(shù):
下載次數(shù):
HTML閱讀次數(shù):
引用次數(shù):

歷史

收稿日期:2023-05-23
最后修改日期:2024-11-25
錄用日期:
在線發(fā)布日期: 2025-11-29
出版日期:

文章二維碼

抽查舔水白紧大视频,冲喜娘子超甜哒激情五月,亚洲黄色图片偷拍网站,亚洲三级黄片免费播放,91在线观看喷潮红桃,日韩精品在线观看一区二区,久久久久久久久久久久黄色片子,日产精品99久久久久久久久,欧洲亚洲另类图片小说

引用本文

分享

相關視頻

文章指標

歷史

文章二維碼

抽查舔水白紧大视频,冲喜娘子超甜哒 激情五月,亚洲黄色图片偷拍网站,亚洲三级黄片免费播放,91在线观看喷潮红桃,日韩精品在线观看一区二区,久久久久久久久久久久黄色片子,日产精品99久久久久久久久,欧洲亚洲另类图片小说

引用本文

分享

相關視頻

文章指標

歷史

文章二維碼

抽查舔水白紧大视频,冲喜娘子超甜哒激情五月,亚洲黄色图片偷拍网站,亚洲三级黄片免费播放,91在线观看喷潮红桃,日韩精品在线观看一区二区,久久久久久久久久久久黄色片子,日产精品99久久久久久久久,欧洲亚洲另类图片小说