Development of predictive models for shallow lake turnovers

Lam, M.Y., Yucel-Bilen, O., Unalan, U.B., Celebi, E.B., Aksoy, A. and Ahmadian, R. ORCID: https://orcid.org/0000-0003-2665-4734 2023. Development of predictive models for shallow lake turnovers. Presented at: 40th IAHR World Congress, 21-25 August 2023. Proceedings of the 40th IAHR World Congress. , vol.40 International Association for Hydro-Environment Engineering and Research, 10.3850/978-90-833476-1-5_iahr40wc-p1631-cd

Full text not available from this repository.

Abstract

Eutrophication, usually caused by excessive nutrients from human activities, may cause dire environmental consequences such as harmful algal blooms and fish kills. Especially, turnover in eutrophic and stratified lakes is dangerous for aquatic lives. Solutions such as reducing storm water run-off and lake restoration are costly and take years to develop. This research develops Artificial neural network (ANN) predictive models for turnovers in Eymir Lake, Turkey with an aim to develop autonomous pre-emptive water quality measurement and intervention system. As there is no consensus as to whether the target variable should be the difference in dissolved oxygen (DO) or temperature between the surface and bottom water layers is better, networks with three separate turnover related target variables, namely (i) difference in dissolved oxygen; (ii) difference in temperature; and (iii) the average of and were trained and compared. Results show high potential for ANN models as all trained networks achieved adjusted R2 over 0.9. The model with the target variable gave best prediction accuracy. However, DO is the one of the direct parameters causing fish kills and other environmental impact, and the selection of target variable should be considered in the context of the purpose of the modelling effort.

Item Type:	Conference or Workshop Item (Paper)
Status:	Published
Schools:	Engineering
Publisher:	International Association for Hydro-Environment Engineering and Research
Last Modified:	10 Apr 2024 15:15
URI:	https://orca.cardiff.ac.uk/id/eprint/167720

Actions (repository staff only)

Edit Item