Nauman Gul , Afed Ullah Khan , Basir Ullah , Bakht Niaz Khan , Hamed M. Almalki , Abdulbasid S. Banga , Kailash Kumar
{"title":"巴基斯坦斯瓦特河流域泥沙负荷预测的LSTM优化:优化器和激活函数的评价","authors":"Nauman Gul , Afed Ullah Khan , Basir Ullah , Bakht Niaz Khan , Hamed M. Almalki , Abdulbasid S. Banga , Kailash Kumar","doi":"10.1016/j.pce.2025.104019","DOIUrl":null,"url":null,"abstract":"<div><div>Accurately estimating sediment load is essential for sustainable water resources management, as excessive sedimentation can reduce reservoir capacity, degrade water quality, and impair aquatic ecosystems, ultimately affecting long-term planning and infrastructure design. However, predictions can be uncertain due to the choice of optimizers and activation functions in machine learning models. This study investigates the influence of seven optimizers (Adam, RMSprop, Adagrad, Adadelta, Adamax, Nadam, and Ftrl) and eight activation functions (ELU, Sigmoid, Linear, Softplus, Swish, SELU, Tanh, and Softmax) on the performance of a Long Short-Term Memory (LSTM) network for sediment load prediction. A total of 56 optimizer-activation combinations were tested using historical data from the Chakdara station in the Swat River basin, Pakistan. The dataset was normalized using Min-Max scaling, with a 70:30 train-test split. Model performance was evaluated using the Coefficient of Determination (R<sup>2</sup>), Mean Squared Error (MSE), Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Percent Bias (PBIAS). Results showed that the Adam-SELU combination achieved the best performance, with an R<sup>2</sup> of 0.81, MSE of 1363.61, and RMSE of 36.93 during training, and an R<sup>2</sup> of 0.80, MSE of 2385.2, and RMSE of 48.84 during testing. This configuration also exhibited minimal bias, with PBIAS values of −1.07 (training) and −1.04 (testing). Other combinations, such as Adam-Nadam and Adam-RMSprop, performed relatively well but with higher error metrics. The findings highlight Adam-SELU as the most effective configuration for sediment load prediction, offering improved accuracy. These insights can guide watershed managers in selecting robust models for sediment forecasting.</div></div>","PeriodicalId":54616,"journal":{"name":"Physics and Chemistry of the Earth","volume":"140 ","pages":"Article 104019"},"PeriodicalIF":3.0000,"publicationDate":"2025-07-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Optimizing LSTM for sediment load prediction in the Swat river basin, Pakistan: Evaluation of optimizers and activation functions\",\"authors\":\"Nauman Gul , Afed Ullah Khan , Basir Ullah , Bakht Niaz Khan , Hamed M. Almalki , Abdulbasid S. Banga , Kailash Kumar\",\"doi\":\"10.1016/j.pce.2025.104019\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><div>Accurately estimating sediment load is essential for sustainable water resources management, as excessive sedimentation can reduce reservoir capacity, degrade water quality, and impair aquatic ecosystems, ultimately affecting long-term planning and infrastructure design. However, predictions can be uncertain due to the choice of optimizers and activation functions in machine learning models. This study investigates the influence of seven optimizers (Adam, RMSprop, Adagrad, Adadelta, Adamax, Nadam, and Ftrl) and eight activation functions (ELU, Sigmoid, Linear, Softplus, Swish, SELU, Tanh, and Softmax) on the performance of a Long Short-Term Memory (LSTM) network for sediment load prediction. A total of 56 optimizer-activation combinations were tested using historical data from the Chakdara station in the Swat River basin, Pakistan. The dataset was normalized using Min-Max scaling, with a 70:30 train-test split. Model performance was evaluated using the Coefficient of Determination (R<sup>2</sup>), Mean Squared Error (MSE), Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Percent Bias (PBIAS). Results showed that the Adam-SELU combination achieved the best performance, with an R<sup>2</sup> of 0.81, MSE of 1363.61, and RMSE of 36.93 during training, and an R<sup>2</sup> of 0.80, MSE of 2385.2, and RMSE of 48.84 during testing. This configuration also exhibited minimal bias, with PBIAS values of −1.07 (training) and −1.04 (testing). Other combinations, such as Adam-Nadam and Adam-RMSprop, performed relatively well but with higher error metrics. The findings highlight Adam-SELU as the most effective configuration for sediment load prediction, offering improved accuracy. These insights can guide watershed managers in selecting robust models for sediment forecasting.</div></div>\",\"PeriodicalId\":54616,\"journal\":{\"name\":\"Physics and Chemistry of the Earth\",\"volume\":\"140 \",\"pages\":\"Article 104019\"},\"PeriodicalIF\":3.0000,\"publicationDate\":\"2025-07-07\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Physics and Chemistry of the Earth\",\"FirstCategoryId\":\"89\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S147470652500169X\",\"RegionNum\":3,\"RegionCategory\":\"地球科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"GEOSCIENCES, MULTIDISCIPLINARY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Physics and Chemistry of the Earth","FirstCategoryId":"89","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S147470652500169X","RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"GEOSCIENCES, MULTIDISCIPLINARY","Score":null,"Total":0}
Optimizing LSTM for sediment load prediction in the Swat river basin, Pakistan: Evaluation of optimizers and activation functions
Accurately estimating sediment load is essential for sustainable water resources management, as excessive sedimentation can reduce reservoir capacity, degrade water quality, and impair aquatic ecosystems, ultimately affecting long-term planning and infrastructure design. However, predictions can be uncertain due to the choice of optimizers and activation functions in machine learning models. This study investigates the influence of seven optimizers (Adam, RMSprop, Adagrad, Adadelta, Adamax, Nadam, and Ftrl) and eight activation functions (ELU, Sigmoid, Linear, Softplus, Swish, SELU, Tanh, and Softmax) on the performance of a Long Short-Term Memory (LSTM) network for sediment load prediction. A total of 56 optimizer-activation combinations were tested using historical data from the Chakdara station in the Swat River basin, Pakistan. The dataset was normalized using Min-Max scaling, with a 70:30 train-test split. Model performance was evaluated using the Coefficient of Determination (R2), Mean Squared Error (MSE), Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Percent Bias (PBIAS). Results showed that the Adam-SELU combination achieved the best performance, with an R2 of 0.81, MSE of 1363.61, and RMSE of 36.93 during training, and an R2 of 0.80, MSE of 2385.2, and RMSE of 48.84 during testing. This configuration also exhibited minimal bias, with PBIAS values of −1.07 (training) and −1.04 (testing). Other combinations, such as Adam-Nadam and Adam-RMSprop, performed relatively well but with higher error metrics. The findings highlight Adam-SELU as the most effective configuration for sediment load prediction, offering improved accuracy. These insights can guide watershed managers in selecting robust models for sediment forecasting.
期刊介绍:
Physics and Chemistry of the Earth is an international interdisciplinary journal for the rapid publication of collections of refereed communications in separate thematic issues, either stemming from scientific meetings, or, especially compiled for the occasion. There is no restriction on the length of articles published in the journal. Physics and Chemistry of the Earth incorporates the separate Parts A, B and C which existed until the end of 2001.
Please note: the Editors are unable to consider submissions that are not invited or linked to a thematic issue. Please do not submit unsolicited papers.
The journal covers the following subject areas:
-Solid Earth and Geodesy:
(geology, geochemistry, tectonophysics, seismology, volcanology, palaeomagnetism and rock magnetism, electromagnetism and potential fields, marine and environmental geosciences as well as geodesy).
-Hydrology, Oceans and Atmosphere:
(hydrology and water resources research, engineering and management, oceanography and oceanic chemistry, shelf, sea, lake and river sciences, meteorology and atmospheric sciences incl. chemistry as well as climatology and glaciology).
-Solar-Terrestrial and Planetary Science:
(solar, heliospheric and solar-planetary sciences, geology, geophysics and atmospheric sciences of planets, satellites and small bodies as well as cosmochemistry and exobiology).