{"id":{"repo_id":"cape-town","oai_identifier":"oai:open.uct.ac.za:11427/43395"},"canonical_url":"https://search.dev.ndltd.org/etd/cape-town/oai:open.uct.ac.za:11427/43395","repository":{"repo_id":"cape-town","name":"University of Cape Town","base_url":"https://open.uct.ac.za/oai/request"},"display":{"title":"Coastal water level prediction: a comparative study of statistical and machine learning techniques for time series forecasting","abstract":"Time series analysis provides powerful tools for predicting future trends, outcomes, and events. The application of these tools to coastal water level forecasting generates insightful predictions for operational use in flood management, as well as a deeper understanding of the influencing factors. Many existing models and projects focus on long-term trends in coastal water levels particularly in terms of climate change and global warming. This project investigated the application of time series analysis with exogenous meteorological variables to the task of generating accurate short-term (≤ 96 hour) forecasts of coastal water levels in a manner that is compatible with real-time monitoring for use in operational flood management. Traditional statistical methods, including regression, autoregressive integrated moving average (ARIMA), and generalised additive models, were compared alongside machine learning methods including extreme gradient boosting, support vector machines, and long short-term memory networks. Extreme gradient boosting with 24-hour of lagged input features was found to have the greatest overall test accuracy and stable predictions over the 96-hour forecast horizon. ARIMA models were the most accurate at predicting water levels in the positive stage (during high-tide). The exogenous meteorological variables contributed significantly to the models' ability to predict the water level.","abstract_html":"Time series analysis provides powerful tools for predicting future trends, outcomes, and events. The application of these tools to coastal water level forecasting generates insightful predictions for operational use in flood management, as well as a deeper understanding of the influencing factors. Many existing models and projects focus on long-term trends in coastal water levels particularly in terms of climate change and global warming. This project investigated the application of time series analysis with exogenous meteorological variables to the task of generating accurate short-term (≤ 96 hour) forecasts of coastal water levels in a manner that is compatible with real-time monitoring for use in operational flood management. Traditional statistical methods, including regression, autoregressive integrated moving average (ARIMA), and generalised additive models, were compared alongside machine learning methods including extreme gradient boosting, support vector machines, and long short-term memory networks. Extreme gradient boosting with 24-hour of lagged input features was found to have the greatest overall test accuracy and stable predictions over the 96-hour forecast horizon. ARIMA models were the most accurate at predicting water levels in the positive stage (during high-tide). The exogenous meteorological variables contributed significantly to the models' ability to predict the water level.","abstract_has_math":false,"creators":["Harrison, Jonathan"],"institution":"Department of Statistical Sciences","degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":["Erni, Birgit","Britz, Stefan","Largier, John"],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026","date_published":"2026","updated_at":"2026-07-24T01:34:03Z","subjects":["coastal water","machine learning"],"languages":["en"],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://hdl.handle.net/11427/43395","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Erni, Birgit","Britz, Stefan","Largier, John"]},{"key":"dc:creator","label":"Author","values":["Harrison, Jonathan"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2026-06-26T07:58:34Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2026-06-26T07:58:34Z"]},{"key":"dc:date.issued","label":"Date","values":["2026"]},{"key":"dc:publisher.department","label":"Dc Publisher Department","values":["Department of Statistical Sciences"]},{"key":"dc:publisher.institution","label":"Dc Publisher Institution","values":["University of Cape Town"]},{"key":"dc:type","label":"Dc Type","values":["Thesis / Dissertation"]},{"key":"dc:type.qualificationlevel","label":"Dc Type Qualificationlevel","values":["Masters","MSc"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["coastal water","machine learning"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["http://hdl.handle.net/11427/43395"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Time series analysis provides powerful tools for predicting future trends, outcomes, and events. The application of these tools to coastal water level forecasting generates insightful predictions for operational use in flood management, as well as a deeper understanding of the influencing factors. Many existing models and projects focus on long-term trends in coastal water levels particularly in terms of climate change and global warming. This project investigated the application of time series analysis with exogenous meteorological variables to the task of generating accurate short-term (≤ 96 hour) forecasts of coastal water levels in a manner that is compatible with real-time monitoring for use in operational flood management. Traditional statistical methods, including regression, autoregressive integrated moving average (ARIMA), and generalised additive models, were compared alongside machine learning methods including extreme gradient boosting, support vector machines, and long short-term memory networks. Extreme gradient boosting with 24-hour of lagged input features was found to have the greatest overall test accuracy and stable predictions over the 96-hour forecast horizon. ARIMA models were the most accurate at predicting water levels in the positive stage (during high-tide). The exogenous meteorological variables contributed significantly to the models' ability to predict the water level."]},{"key":"dc:title","label":"Title","values":["Coastal water level prediction: a comparative study of statistical and machine learning techniques for time series forecasting"]}]}],"canonical_facts":{"dc:contributor.advisor":["Erni, Birgit","Britz, Stefan","Largier, John"],"dc:creator":["Harrison, Jonathan"],"dc:date.accessioned":["2026-06-26T07:58:34Z"],"dc:date.available":["2026-06-26T07:58:34Z"],"dc:date.issued":["2026"],"dc:description.abstract":["Time series analysis provides powerful tools for predicting future trends, outcomes, and events. The application of these tools to coastal water level forecasting generates insightful predictions for operational use in flood management, as well as a deeper understanding of the influencing factors. Many existing models and projects focus on long-term trends in coastal water levels particularly in terms of climate change and global warming. This project investigated the application of time series analysis with exogenous meteorological variables to the task of generating accurate short-term (≤ 96 hour) forecasts of coastal water levels in a manner that is compatible with real-time monitoring for use in operational flood management. Traditional statistical methods, including regression, autoregressive integrated moving average (ARIMA), and generalised additive models, were compared alongside machine learning methods including extreme gradient boosting, support vector machines, and long short-term memory networks. Extreme gradient boosting with 24-hour of lagged input features was found to have the greatest overall test accuracy and stable predictions over the 96-hour forecast horizon. ARIMA models were the most accurate at predicting water levels in the positive stage (during high-tide). The exogenous meteorological variables contributed significantly to the models' ability to predict the water level."],"dc:identifier.uri":["http://hdl.handle.net/11427/43395"],"dc:language.iso":["en"],"dc:publisher.department":["Department of Statistical Sciences"],"dc:publisher.institution":["University of Cape Town"],"dc:subject":["coastal water","machine learning"],"dc:title":["Coastal water level prediction: a comparative study of statistical and machine learning techniques for time series forecasting"],"dc:type":["Thesis / Dissertation"],"dc:type.qualificationlevel":["Masters","MSc"]},"updated_at":"2026-07-24T01:34:03Z"}