Scholarlink

<p>Drought is a significant disaster affecting human life on multiple levels, including a decrease in crop yield, desertification, socio-economic effects, and human health. Replicating past scenarios and forecasting them for the future is crucial to formulate policies that can minimize the impacts of drought. This study aims to assess hydrological drought conditions by reconstructing historical drought patterns and evaluating model performance for accurate forecasting. The research used the Soil and Water Assessment Tool (SWAT) hydrological model and deep learning-based models such as Long Short-Term Memory (LSTM), Bidirectional LSTM (BiLSTM), Gated Recurrent Unit (GRU), and Convolutional-LSTM (Conv-LSTM) to reconstruct historical conditions. For the assessment of hydrological drought in the Divandareh-Bijar basin, which is located in Iran, the study employed the 1, 3, 6, and 12-month Standardized Streamflow Index (SSI) as drought indices between 1988 and 2010. Trend and change-point analyses were conducted using the non-parametric Mann-Kendall and Pettitt tests to evaluate temporal variability in streamflow. The findings indicated that the SWAT model was better than other proposed models in the reconstruction of past conditions based on measures such as Nash-Sutcliffe Efficiency (NSE), Percentage Bias (Pbias), and Root Mean Squared Error-observations Standard deviation Ratio (RSR). The SWAT model was found to be 14% superior to AI-based models with an NSE of 0.8, Pbias of 8.8, and RSR of 0.4. Furthermore, the SSI analysis demonstrated the presence of both wet and dry spells throughout the study period, and revealed that the climate change impact rate was higher than the anthropogenic activities by 76% and 24%, respectively. In conclusion, the SWAT model proved more effective for drought reconstruction, and the results emphasize the dominant role of climate change in driving hydrological drought in the study area.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Reconstructing Hydrological Drought Using SWAT, Deep Learning Models, and SSI

  • 期刊论文

  • 发表日期：

    2026年3月28日

查看全文
  • Vahid Nourani,
  • Habibeh Abbasi,
  • Arman Hosseinpour Salehi,
  • Mohammad Taghi Aalami,
  • Mohammad Bejani,
  • Hadi Pourali

摘要

Drought is a significant disaster affecting human life on multiple levels, including a decrease in crop yield, desertification, socio-economic effects, and human health. Replicating past scenarios and forecasting them for the future is crucial to formulate policies that can minimize the impacts of drought. This study aims to assess hydrological drought conditions by reconstructing historical drought patterns and evaluating model performance for accurate forecasting. The research used the Soil and Water Assessment Tool (SWAT) hydrological model and deep learning-based models such as Long Short-Term Memory (LSTM), Bidirectional LSTM (BiLSTM), Gated Recurrent Unit (GRU), and Convolutional-LSTM (Conv-LSTM) to reconstruct historical conditions. For the assessment of hydrological drought in the Divandareh-Bijar basin, which is located in Iran, the study employed the 1, 3, 6, and 12-month Standardized Streamflow Index (SSI) as drought indices between 1988 and 2010. Trend and change-point analyses were conducted using the non-parametric Mann-Kendall and Pettitt tests to evaluate temporal variability in streamflow. The findings indicated that the SWAT model was better than other proposed models in the reconstruction of past conditions based on measures such as Nash-Sutcliffe Efficiency (NSE), Percentage Bias (Pbias), and Root Mean Squared Error-observations Standard deviation Ratio (RSR). The SWAT model was found to be 14% superior to AI-based models with an NSE of 0.8, Pbias of 8.8, and RSR of 0.4. Furthermore, the SSI analysis demonstrated the presence of both wet and dry spells throughout the study period, and revealed that the climate change impact rate was higher than the anthropogenic activities by 76% and 24%, respectively. In conclusion, the SWAT model proved more effective for drought reconstruction, and the results emphasize the dominant role of climate change in driving hydrological drought in the study area.