Storm surges cause coastal flooding, one of the most devastating coastal hazards. Accurate modeling of storm surges is essential for predicting and mitigating these impacts. Traditional approaches model surges at individual tide gauges and often focus on daily time scales, leading to data redundancy and limiting their ability to capture sub-daily variability. This study addresses these limitations using deep learning (DL) algorithms to model hourly surges simultaneously at multiple tide gauges along the U.S. East Coast and Gulf of Mexico. Three algorithms, Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and a hybrid ConvLSTM are employed to model surges using atmospheric variables (e.g., sea level pressure and winds) as predictors. ConvLSTM outperforms the others in predicting the overall variability of surges, extreme surge events, and statistical attributes of extreme surge hydrographs. The model’s efficiency is comparable to that of existing data-driven and hydrodynamic models in reproducing surges. The study also applies DL models to project future changes in storm surge statistics, using predictor variables from the GFDL-ESM4 global climate model (GCM) under two climate scenarios. In general, there will be increases in the mean and extreme surges at some TGs, whereas the others will experience a reduction. However, it is difficult to draw robust conclusive remarks on the future changes in storm surges as the study considers only one GCM. Therefore, considering more GCMs to cover the whole range of uncertainty and computing the uncertainty in the future changes of storm surge statistics is crucial and proposed as a future extension of this research.
","abstract_html":"<p>Storm surges cause coastal flooding, one of the most devastating coastal hazards. Accurate modeling of storm surges is essential for predicting and mitigating these impacts. Traditional approaches model surges at individual tide gauges and often focus on daily time scales, leading to data redundancy and limiting their ability to capture sub-daily variability. This study addresses these limitations using deep learning (DL) algorithms to model hourly surges simultaneously at multiple tide gauges along the U.S. East Coast and Gulf of Mexico. Three algorithms, Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and a hybrid ConvLSTM are employed to model surges using atmospheric variables (e.g., sea level pressure and winds) as predictors. ConvLSTM outperforms the others in predicting the overall variability of surges, extreme surge events, and statistical attributes of extreme surge hydrographs. The model’s efficiency is comparable to that of existing data-driven and hydrodynamic models in reproducing surges. The study also applies DL models to project future changes in storm surge statistics, using predictor variables from the GFDL-ESM4 global climate model (GCM) under two climate scenarios. In general, there will be increases in the mean and extreme surges at some TGs, whereas the others will experience a reduction. However, it is difficult to draw robust conclusive remarks on the future changes in storm surges as the study considers only one GCM. Therefore, considering more GCMs to cover the whole range of uncertainty and computing the uncertainty in the future changes of storm surge statistics is crucial and proposed as a future extension of this research.</p>","abstract_has_math":false,"creators":["Zafor, Md Abu"],"institution":null,"degree_name":"Master of Science (MS)","degree_level":"Masters Thesis","degree_discipline":null,"degree_department":null,"school":null,"contributors":["Dr. Md Mamunur Rashid","Dr.Wei Wu","Dr.Robert Leaf"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2024,"date_issued":"2024-12-01T08:00:00Z","date_published":"2024-12-01T08:00:00Z","updated_at":"2026-07-24T05:45:47Z","subjects":["Storm Surge; Sea Level; Deep Learning; CNN; LSTM; Convo-LSTM; CMIP6; GCMs","Atmospheric Sciences","Climate","Meteorology","Oceanography","Other Oceanography and Atmospheric Sciences and Meteorology"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://aquila.usm.edu/masters_theses/1082","outbound_label":"Repository record","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Dr. Md Mamunur Rashid","Dr.Wei Wu","Dr.Robert Leaf"]},{"key":"dc:creator","label":"Author","values":["Zafor, Md Abu"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.available","label":"Dc Date Available","values":["2026-07-01T07:00:00Z"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Masters Thesis"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Master of Science (MS)"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Storm Surge; Sea Level; Deep Learning; CNN; LSTM; Convo-LSTM; CMIP6; GCMs","Atmospheric Sciences","Climate","Meteorology","Oceanography","Other Oceanography and Atmospheric Sciences and Meteorology"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://aquila.usm.edu/masters_theses/1082"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Storm surges cause coastal flooding, one of the most devastating coastal hazards. Accurate modeling of storm surges is essential for predicting and mitigating these impacts. Traditional approaches model surges at individual tide gauges and often focus on daily time scales, leading to data redundancy and limiting their ability to capture sub-daily variability. This study addresses these limitations using deep learning (DL) algorithms to model hourly surges simultaneously at multiple tide gauges along the U.S. East Coast and Gulf of Mexico. Three algorithms, Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and a hybrid ConvLSTM are employed to model surges using atmospheric variables (e.g., sea level pressure and winds) as predictors. ConvLSTM outperforms the others in predicting the overall variability of surges, extreme surge events, and statistical attributes of extreme surge hydrographs. The model’s efficiency is comparable to that of existing data-driven and hydrodynamic models in reproducing surges. The study also applies DL models to project future changes in storm surge statistics, using predictor variables from the GFDL-ESM4 global climate model (GCM) under two climate scenarios. In general, there will be increases in the mean and extreme surges at some TGs, whereas the others will experience a reduction. However, it is difficult to draw robust conclusive remarks on the future changes in storm surges as the study considers only one GCM. Therefore, considering more GCMs to cover the whole range of uncertainty and computing the uncertainty in the future changes of storm surge statistics is crucial and proposed as a future extension of this research.
"]},{"key":"dc:title","label":"Title","values":["Modeling Hourly Storm Surges using Deep Learning Techniques"]}]}],"canonical_facts":{"dc:contributor":["Dr. Md Mamunur Rashid","Dr.Wei Wu","Dr.Robert Leaf"],"dc:creator":["Zafor, Md Abu"],"dc:date.available":["2026-07-01T07:00:00Z"],"dc:description.abstract":["Storm surges cause coastal flooding, one of the most devastating coastal hazards. Accurate modeling of storm surges is essential for predicting and mitigating these impacts. Traditional approaches model surges at individual tide gauges and often focus on daily time scales, leading to data redundancy and limiting their ability to capture sub-daily variability. This study addresses these limitations using deep learning (DL) algorithms to model hourly surges simultaneously at multiple tide gauges along the U.S. East Coast and Gulf of Mexico. Three algorithms, Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and a hybrid ConvLSTM are employed to model surges using atmospheric variables (e.g., sea level pressure and winds) as predictors. ConvLSTM outperforms the others in predicting the overall variability of surges, extreme surge events, and statistical attributes of extreme surge hydrographs. The model’s efficiency is comparable to that of existing data-driven and hydrodynamic models in reproducing surges. The study also applies DL models to project future changes in storm surge statistics, using predictor variables from the GFDL-ESM4 global climate model (GCM) under two climate scenarios. In general, there will be increases in the mean and extreme surges at some TGs, whereas the others will experience a reduction. However, it is difficult to draw robust conclusive remarks on the future changes in storm surges as the study considers only one GCM. Therefore, considering more GCMs to cover the whole range of uncertainty and computing the uncertainty in the future changes of storm surge statistics is crucial and proposed as a future extension of this research.
"],"dc:identifier":["https://aquila.usm.edu/masters_theses/1082"],"dc:subject":["Storm Surge; Sea Level; Deep Learning; CNN; LSTM; Convo-LSTM; CMIP6; GCMs","Atmospheric Sciences","Climate","Meteorology","Oceanography","Other Oceanography and Atmospheric Sciences and Meteorology"],"dc:title":["Modeling Hourly Storm Surges using Deep Learning Techniques"],"thesis:degree_level":["Masters Thesis"],"thesis:degree_name":["Master of Science (MS)"]},"updated_at":"2026-07-24T05:45:47Z"}