Uncertainties in groundwater-surface water modelling for the HOBE catchment
Publikation: Bog/antologi/afhandling/rapport › Ph.d.-afhandling › Forskning
Standard
Uncertainties in groundwater-surface water modelling for the HOBE catchment. / Ehlers, Lennart Benjamin.
Department of Geosciences and Natural Resource Management, Faculty of Science, University of Copenhagen, 2018. 195 s.Publikation: Bog/antologi/afhandling/rapport › Ph.d.-afhandling › Forskning
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - BOOK
T1 - Uncertainties in groundwater-surface water modelling for the HOBE catchment
AU - Ehlers, Lennart Benjamin
PY - 2018/4
Y1 - 2018/4
N2 - The objective of this Ph.D. thesis was to quantify uncertainties associated with a complex coupled subsurface-surface-atmosphere model. Special focus was on an adequate representation of spatial variability in natural hydrological processes. In contrast to the majority of studies dealing with uncertainty analysis, uncertainties were not only assessed for the lumped catchment response discharge, but for the hydrological variables hydraulic head, soil moisture, and actual evapotranspiration at locations within the study area. In recognition of the specific requirements of the hydrological model, emphasis was put on employing statistical methods characterized by computational efficiency. The Ph.D. study was part of the HOBE (Hydrological Observatory) project established in the Skjern river catchment (2500 km2) in Western Denmark in 2007. The thesis benefited from abundant observational data collected in its course. The main study area of the thesis was the Ahlergaarde (1055 km2) subcatchment. The hydrological model was set up using the MIKE SHE code and largely corresponded to the Danish National Hydrological Model (DK-Model). The thesis is intended to contribute to an improved understanding and treatment of uncertainties in hydrological modelling and further establish uncertainty analysis in operational hydrology. Overall, the following main findings were obtained: The joint treatment of several sources of rainfall uncertainty led to higher overall rainfall uncertainty for all investigated spatial and temporal supports. This was especially true for individual grid locations and is thus relevant when looking at hydrological states and fluxes simulated at interior catchment locations. The incorporation of neighborhood uncertainty had the largest impact on long-term averages of rainfall.The comparison of these results with those obtained by expert elicitation suggested that the estimated rainfall uncertainty is still too low. Based on the evaluation of the hydrological variables discharge, hydraulic head, soil moisture, and actual evapotranspiration at several locations, rainfall uncertainty was found to be more important for overall predictive uncertainty of the investigated subsurface-surface-atmosphere model than parameter uncertainty. The consideration of effective observational uncertainty to account for uncertainties in observations and addressing scale issues allowed for a more realistic performance assessment of the uncertainty analysis. However, the approach exhibited limitations in the presence of strong model biases. The application of a post-processing technique (k-nearest neighbor resampling) to obtain estimates of residual uncertainty was found to be an efficient and robust tool to perform uncertainty analysis. Successful application to a computationally heavy subsurface-surface-atmosphere model suggested a large potential of the method for an application in operational hydrology: obtained results showed highly satisfactory coverage of observations by the prediction intervals, while the reliability of the residual uncertainty estimation was illustrated by reliability diagrams and alpha index values. Corroborating results from previous studies, error characteristics were successfully reproduced and biases (thus implicitly considering model structural uncertainty) corrected for.
AB - The objective of this Ph.D. thesis was to quantify uncertainties associated with a complex coupled subsurface-surface-atmosphere model. Special focus was on an adequate representation of spatial variability in natural hydrological processes. In contrast to the majority of studies dealing with uncertainty analysis, uncertainties were not only assessed for the lumped catchment response discharge, but for the hydrological variables hydraulic head, soil moisture, and actual evapotranspiration at locations within the study area. In recognition of the specific requirements of the hydrological model, emphasis was put on employing statistical methods characterized by computational efficiency. The Ph.D. study was part of the HOBE (Hydrological Observatory) project established in the Skjern river catchment (2500 km2) in Western Denmark in 2007. The thesis benefited from abundant observational data collected in its course. The main study area of the thesis was the Ahlergaarde (1055 km2) subcatchment. The hydrological model was set up using the MIKE SHE code and largely corresponded to the Danish National Hydrological Model (DK-Model). The thesis is intended to contribute to an improved understanding and treatment of uncertainties in hydrological modelling and further establish uncertainty analysis in operational hydrology. Overall, the following main findings were obtained: The joint treatment of several sources of rainfall uncertainty led to higher overall rainfall uncertainty for all investigated spatial and temporal supports. This was especially true for individual grid locations and is thus relevant when looking at hydrological states and fluxes simulated at interior catchment locations. The incorporation of neighborhood uncertainty had the largest impact on long-term averages of rainfall.The comparison of these results with those obtained by expert elicitation suggested that the estimated rainfall uncertainty is still too low. Based on the evaluation of the hydrological variables discharge, hydraulic head, soil moisture, and actual evapotranspiration at several locations, rainfall uncertainty was found to be more important for overall predictive uncertainty of the investigated subsurface-surface-atmosphere model than parameter uncertainty. The consideration of effective observational uncertainty to account for uncertainties in observations and addressing scale issues allowed for a more realistic performance assessment of the uncertainty analysis. However, the approach exhibited limitations in the presence of strong model biases. The application of a post-processing technique (k-nearest neighbor resampling) to obtain estimates of residual uncertainty was found to be an efficient and robust tool to perform uncertainty analysis. Successful application to a computationally heavy subsurface-surface-atmosphere model suggested a large potential of the method for an application in operational hydrology: obtained results showed highly satisfactory coverage of observations by the prediction intervals, while the reliability of the residual uncertainty estimation was illustrated by reliability diagrams and alpha index values. Corroborating results from previous studies, error characteristics were successfully reproduced and biases (thus implicitly considering model structural uncertainty) corrected for.
UR - https://soeg.kb.dk/permalink/45KBDK_KGL/1pioq0f/alma99122584242405763
M3 - Ph.D. thesis
BT - Uncertainties in groundwater-surface water modelling for the HOBE catchment
PB - Department of Geosciences and Natural Resource Management, Faculty of Science, University of Copenhagen
ER -
ID: 197467600