Pregled bibliografske jedinice broj: 1185519
Implementation of a Long Short-Term Memory Neural Network based hydrological model in a snow dominated Alpine basin
Implementation of a Long Short-Term Memory Neural Network based hydrological model in a snow dominated Alpine basin // XXIX Conference of the Danubian Countries on Hydrological Forecasting and Hydrological Bases of Water Management
Prag: Czech Hydrometeorological Institute, 2021. str. 75-77 (predavanje, međunarodna recenzija, prošireni sažetak, znanstveni)
CROSBI ID: 1185519 Za ispravke kontaktirajte CROSBI podršku putem web obrasca
Naslov
Implementation of a Long Short-Term Memory Neural
Network based hydrological model in a snow
dominated Alpine basin
Autori
Leskovar, Karlo ; Bekić, Damir ; Težak, Denis ; Meaški, Hrvoje
Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, prošireni sažetak, znanstveni
Izvornik
XXIX Conference of the Danubian Countries on Hydrological Forecasting and Hydrological Bases of Water Management
/ - Prag : Czech Hydrometeorological Institute, 2021, 75-77
ISBN
978-80-7653-020-1
Skup
XXIX Conference of the Danubian Countries on hydrological Forecasting and Hydrological Bases of Water Managemen
Mjesto i datum
Prag, Češka Republika, 06.09.2021. - 08.09.2021
Vrsta sudjelovanja
Predavanje
Vrsta recenzije
Međunarodna recenzija
Ključne riječi
hydrological model ; snow dominated basin ; Artificial Neural Network ; Long Short-Term memory ; Drava river
Sažetak
Snow dominated Alpine basins are of great importance for the surrounding areas. Due to complex terrain and weather phenomena, hydrological modelling in these areas is often difficult. In this article a state-of-the-art approach based on data-driven models, in form of Recurrent Artificial Neural Networks is presented. In order to investigate the influence of different input data types three models were created. The results proved that adding more input data features can improve model prediction capabilities (R2 increased from 0.91 to 0.93 for the testing period), although fine tuning of model hyperparameters is mandatory to achieve such results. The results of this study show that Long Short-term memory neural network based hydrological models can be used in mid to high elevation snowmelt dominated basins of the Danube tributaries.
Izvorni jezik
Engleski
Znanstvena područja
Građevinarstvo, Računarstvo, Interdisciplinarne tehničke znanosti
POVEZANOST RADA
Ustanove:
Građevinski fakultet, Zagreb,
Geotehnički fakultet, Varaždin