Characteristics and prediction of extreme drought event using LSTM model in Wei River Basin

As one of the natural disasters, drought has caused a large amount of financial loss in the past few centuries. It is quite essential to reveal the variation of extreme drought event in Wei River Basin (WRB) of China. This paper investigated the change patterns of extreme drought event using Standar...

Full description

Bibliographic Details
Main Authors: Dongfei Yan, Rengui Jiang, Jiancang Xie, Yong Zhao, Jiwei Zhu, Jichao Liang
Format: Article
Language:English
Published: Chinese Geoscience Union 2021-04-01
Series:Terrestrial, Atmospheric and Oceanic Sciences
Online Access: http://tao.cgu.org.tw/media/k2/attachments/v322p261.pdf
id doaj-3fd51be4848647a3b74040e0d9e8d662
record_format Article
spelling doaj-3fd51be4848647a3b74040e0d9e8d6622021-07-26T08:14:36ZengChinese Geoscience UnionTerrestrial, Atmospheric and Oceanic Sciences1017-08392311-76802021-04-0132226127410.3319/TAO.2021.02.07.01Characteristics and prediction of extreme drought event using LSTM model in Wei River BasinDongfei YanRengui JiangJiancang XieYong ZhaoJiwei ZhuJichao LiangAs one of the natural disasters, drought has caused a large amount of financial loss in the past few centuries. It is quite essential to reveal the variation of extreme drought event in Wei River Basin (WRB) of China. This paper investigated the change patterns of extreme drought event using Standardized Precipitation Evapotranspiration Index (SPEI). Furthermore, the SPEI is predicted by combining different influencing factors in the WRB using Long Short-Term Memory (LSTM) model. The spatiotemporal variation characteristics were examined using non-parametric Mann-Kendall test, and the nonlinear relationships between El Niño-Southern Oscillation (ENSO) and SPEI were quantified using wavelet coherence analysis (WTC). Results showed that midland of the WRB have the highest probability of extreme drought events. Meanwhile, changes in SPEI in the northeast were more erratic than in other regions. The area with extreme drought had increased at a rate of 2.4% per decade. The prediction result of SPEI-24 was the best by LSTM model, and the prediction result of SPEI-3 was the worst. The R-square between the predicted value and the actual value of SPEI-24 is 0.87. The results help to realize the characteristics of extreme drought in the last hundreds of years, which can provide scientific basis and reference for drought emergency response and management in the WRB. http://tao.cgu.org.tw/media/k2/attachments/v322p261.pdf
collection DOAJ
language English
format Article
sources DOAJ
author Dongfei Yan
Rengui Jiang
Jiancang Xie
Yong Zhao
Jiwei Zhu
Jichao Liang
spellingShingle Dongfei Yan
Rengui Jiang
Jiancang Xie
Yong Zhao
Jiwei Zhu
Jichao Liang
Characteristics and prediction of extreme drought event using LSTM model in Wei River Basin
Terrestrial, Atmospheric and Oceanic Sciences
author_facet Dongfei Yan
Rengui Jiang
Jiancang Xie
Yong Zhao
Jiwei Zhu
Jichao Liang
author_sort Dongfei Yan
title Characteristics and prediction of extreme drought event using LSTM model in Wei River Basin
title_short Characteristics and prediction of extreme drought event using LSTM model in Wei River Basin
title_full Characteristics and prediction of extreme drought event using LSTM model in Wei River Basin
title_fullStr Characteristics and prediction of extreme drought event using LSTM model in Wei River Basin
title_full_unstemmed Characteristics and prediction of extreme drought event using LSTM model in Wei River Basin
title_sort characteristics and prediction of extreme drought event using lstm model in wei river basin
publisher Chinese Geoscience Union
series Terrestrial, Atmospheric and Oceanic Sciences
issn 1017-0839
2311-7680
publishDate 2021-04-01
description As one of the natural disasters, drought has caused a large amount of financial loss in the past few centuries. It is quite essential to reveal the variation of extreme drought event in Wei River Basin (WRB) of China. This paper investigated the change patterns of extreme drought event using Standardized Precipitation Evapotranspiration Index (SPEI). Furthermore, the SPEI is predicted by combining different influencing factors in the WRB using Long Short-Term Memory (LSTM) model. The spatiotemporal variation characteristics were examined using non-parametric Mann-Kendall test, and the nonlinear relationships between El Niño-Southern Oscillation (ENSO) and SPEI were quantified using wavelet coherence analysis (WTC). Results showed that midland of the WRB have the highest probability of extreme drought events. Meanwhile, changes in SPEI in the northeast were more erratic than in other regions. The area with extreme drought had increased at a rate of 2.4% per decade. The prediction result of SPEI-24 was the best by LSTM model, and the prediction result of SPEI-3 was the worst. The R-square between the predicted value and the actual value of SPEI-24 is 0.87. The results help to realize the characteristics of extreme drought in the last hundreds of years, which can provide scientific basis and reference for drought emergency response and management in the WRB.
url http://tao.cgu.org.tw/media/k2/attachments/v322p261.pdf
work_keys_str_mv AT dongfeiyan characteristicsandpredictionofextremedroughteventusinglstmmodelinweiriverbasin
AT renguijiang characteristicsandpredictionofextremedroughteventusinglstmmodelinweiriverbasin
AT jiancangxie characteristicsandpredictionofextremedroughteventusinglstmmodelinweiriverbasin
AT yongzhao characteristicsandpredictionofextremedroughteventusinglstmmodelinweiriverbasin
AT jiweizhu characteristicsandpredictionofextremedroughteventusinglstmmodelinweiriverbasin
AT jichaoliang characteristicsandpredictionofextremedroughteventusinglstmmodelinweiriverbasin
_version_ 1721282075585150976