Filtern
Volltext vorhanden
- nein (2)
Erscheinungsjahr
- 2021 (2)
Dokumenttyp
- Wissenschaftlicher Artikel (2) (entfernen)
Sprache
- Englisch (2) (entfernen)
Gehört zur Bibliographie
- ja (2)
Schlagworte
- modelling (2) (entfernen)
Institut
- Institut für Umweltwissenschaften und Geographie (2) (entfernen)
We systematically explore the effect of calibration data length on the performance of a conceptual hydrological model, GR4H, in comparison to two Artificial Neural Network (ANN) architectures: Long Short-Term Memory Networks (LSTM) and Gated Recurrent Units (GRU), which have just recently been introduced to the field of hydrology. We implemented a case study for six river basins across the contiguous United States, with 25 years of meteorological and discharge data. Nine years were reserved for independent validation; two years were used as a warm-up period, one year for each of the calibration and validation periods, respectively; from the remaining 14 years, we sampled increasing amounts of data for model calibration, and found pronounced differences in model performance. While GR4H required less data to converge, LSTM and GRU caught up at a remarkable rate, considering their number of parameters. Also, LSTM and GRU exhibited the higher calibration instability in comparison to GR4H. These findings confirm the potential of modern deep-learning architectures in rainfall runoff modelling, but also highlight the noticeable differences between them in regard to the effect of calibration data length.
Over the past decades, floods have caused significant financial losses in Turkey, amounting to US$ 800 million between 1960 and 2014. With the Sendai Framework for Disaster Risk Reduction 2015-2030 (SFDRR), it is aimed to reduce the direct economic loss from disasters in relation to the global gross domestic product (GDP) by 2030. Accordingly, a methodology based on experiences from developing countries was proposed by the United Nations Office for Disaster Risk Reduction (UNDRR) to estimate direct economic losses on the macro-scale. Since Turkey also signed the SFDRR, we aimed to adapt, validate and apply the loss estimation model proposed by the UNDRR in Turkey for the first time. To do so, the well-documented flood event in Mersin of 2016 was used to calibrate the damage ratios for the agricultural, commercial and residential sectors, as well as educational facilities. Case studies between 2015 and 2020 with documented losses were further used to validate the model. Finally, model applications provided initial loss estimates for floods occurred recently in Turkey. Despite the limited event documentation for each sector, the calibrated model yielded good results when compared to documented losses. Thus, by implementing the UNDRR method, this study provides an approach to estimate the direct economic losses in Turkey on the macro-scale, which can be used to fill gaps in event databases, support the coordination of financial aid after flood events and facilitate monitoring of the progress toward and achievement of Global Target C of the Sendai Framework for Disaster Risk Reduction 2015-2030.