Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Short-term sea level forecasting using machine learning techniques: A case study for South Africa
Seawater levels along the South African coastline are investigated with the use of machine learning techniques. In this study, data-driven methods, which are more computationally efficient in comparison to numerical models, are applied to predict seawater levels. The open-loop NARX model was develop...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | English |
| Published: |
Department of Oceanography
2023
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867613168079994880 |
|---|---|
| access_status_str | Open Access |
| author | Ingreso, Maria Kristina |
| author2 | Smit, Albertus |
| author_browse | Ingreso, Maria Kristina Smit, Albertus |
| author_facet | Smit, Albertus Ingreso, Maria Kristina |
| author_sort | Ingreso, Maria Kristina |
| collection | Thesis |
| description | Seawater levels along the South African coastline are investigated with the use of machine learning techniques. In this study, data-driven methods, which are more computationally efficient in comparison to numerical models, are applied to predict seawater levels. The open-loop NARX model was developed using the Neural Net Time Series application from the Deep Learning Toolbox 14.0 provided by MATLAB® (Mathworks, 2020). A total of five inputs (atmospheric pressure, mean wave period and direction, wind speed and direction) and a single output of seawater level was fed into the neural network where 70 % of the data was used for training, 15 % was used for validation and the remaining 15 % was used to test the model. Three separate storm events that occurred along the coast of South Africa were used for the final model validation. Model performance was measured using the correlation coefficient (R), the root mean square error (RMSE), the bias and the Willmott indices of correlation. It was found that, through principal component analysis (PCA), atmospheric pressure, wind speed and direction and mean wave period and direction are important physical drivers of sea level. The overall model performance was better when all five met-ocean variables were included as inputs to the model than when one or two were excluded, with R and RMSE values ranging from 0.85 to 0.99 and 4.344 to 100.5 mm, respectively. The study presented here clearly shows an effective methodology to not only demonstrate the high accuracy the model has on seawater level predictions, but also able to further investigate the importance of what each oceanic and atmospheric variable has on the seawater level. The model performance may be affected by frictional shoaling, coastally trapped waves, bathymetry and the local dynamics contributed by Agulhas Current, which were not taken account for in this study and could be incorporated in the model for future research. |
| format | Thesis |
| id | oai:open.uct.ac.za:11427/37386 |
| institution | University of Cape Town (South Africa) |
| language | eng |
| last_indexed | 2026-06-10T12:31:50.330Z |
| license_str | Not specified — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from UCTD — University of Cape Town Open Access Repository |
| publishDate | 2023 |
| publishDateRange | 2023 |
| publishDateSort | 2023 |
| publisher | Department of Oceanography |
| publisherStr | Department of Oceanography |
| record_format | dspace |
| source_str | UCTD — University of Cape Town Open Access Repository |
| spelling | oai:open.uct.ac.za:11427/37386 Short-term sea level forecasting using machine learning techniques: A case study for South Africa Ingreso, Maria Kristina Smit, Albertus Rautenbach, Christo Vichi, Marcello Oceanography Seawater levels along the South African coastline are investigated with the use of machine learning techniques. In this study, data-driven methods, which are more computationally efficient in comparison to numerical models, are applied to predict seawater levels. The open-loop NARX model was developed using the Neural Net Time Series application from the Deep Learning Toolbox 14.0 provided by MATLAB® (Mathworks, 2020). A total of five inputs (atmospheric pressure, mean wave period and direction, wind speed and direction) and a single output of seawater level was fed into the neural network where 70 % of the data was used for training, 15 % was used for validation and the remaining 15 % was used to test the model. Three separate storm events that occurred along the coast of South Africa were used for the final model validation. Model performance was measured using the correlation coefficient (R), the root mean square error (RMSE), the bias and the Willmott indices of correlation. It was found that, through principal component analysis (PCA), atmospheric pressure, wind speed and direction and mean wave period and direction are important physical drivers of sea level. The overall model performance was better when all five met-ocean variables were included as inputs to the model than when one or two were excluded, with R and RMSE values ranging from 0.85 to 0.99 and 4.344 to 100.5 mm, respectively. The study presented here clearly shows an effective methodology to not only demonstrate the high accuracy the model has on seawater level predictions, but also able to further investigate the importance of what each oceanic and atmospheric variable has on the seawater level. The model performance may be affected by frictional shoaling, coastally trapped waves, bathymetry and the local dynamics contributed by Agulhas Current, which were not taken account for in this study and could be incorporated in the model for future research. 2023-03-13T11:17:23Z 2023-03-13T11:17:23Z 2022 2023-02-20T12:57:40Z Master Thesis Masters MSc http://hdl.handle.net/11427/37386 eng application/pdf Department of Oceanography Faculty of Science |
| spellingShingle | Oceanography Ingreso, Maria Kristina Short-term sea level forecasting using machine learning techniques: A case study for South Africa |
| thesis_degree_str | Master's |
| title | Short-term sea level forecasting using machine learning techniques: A case study for South Africa |
| title_full | Short-term sea level forecasting using machine learning techniques: A case study for South Africa |
| title_fullStr | Short-term sea level forecasting using machine learning techniques: A case study for South Africa |
| title_full_unstemmed | Short-term sea level forecasting using machine learning techniques: A case study for South Africa |
| title_short | Short-term sea level forecasting using machine learning techniques: A case study for South Africa |
| title_sort | short term sea level forecasting using machine learning techniques a case study for south africa |
| topic | Oceanography |
| url | http://hdl.handle.net/11427/37386 |
| work_keys_str_mv | AT ingresomariakristina shorttermsealevelforecastingusingmachinelearningtechniquesacasestudyforsouthafrica |