In:
Österreichische Wasser- und Abfallwirtschaft, Springer Science and Business Media LLC, Vol. 74, No. 11-12 ( 2022-12), p. 469-485
Abstract:
This study comprises the prediction of runoff characteristics for high water (MJHQ), mean water (MQ), and low water (MJNQ, MJNQ 7 , Q 95 , Q 98 ) for all topografic catchments of the Austrian surface water bodies including the foreign hydrological upstream regions. The machine learning model XGBoost was applied for the regionalization of the six runoff characteristics. The LamaH dataset was used for training XGBoost, which includes over 70 aggregated catchment characteristics and 15 meteorological time series for 859 observed catchments in Central Europe. Anthropogenic influences such as reservoirs or cross-basin water transfers were considered in the model by additionally created attributes. The test results showed that a deviation of approximately 20% can be expected for the prediction of runoff characteristics in ungauged catchments, which also includes highly anthropogenically influenced catchments. Furthermore, the 90% confidence interval of each prediction was estimated and classified using a Quantile Random Forest model. The results are provided free of charge to the public in form of shapefiles at https://doi.org/10.5281/zenodo.6523372
Type of Medium:
Online Resource
ISSN:
0945-358X
,
1613-7566
DOI:
10.1007/s00506-022-00891-4
Language:
German
Publisher:
Springer Science and Business Media LLC
Publication Date:
2022
detail.hit.zdb_id:
1186984-7
detail.hit.zdb_id:
2383304-X
Permalink