TY - BOOK

T1 - Data Assimilation in Integrated and Distributed Hydrological Models

AU - Zhang, Donghua

N1 - Ph.d.-grad opnået ved mundtligt forsvar 20. maj 2016

PY - 2016/4/8

Y1 - 2016/4/8

N2 - Integrated hydrological models are frequently used in water-related environmental resource management. With our better understanding of the hydrological processes and the improved computational power, hydrological models are becoming increasingly more complex as they integrate multiple hydrological processes and provide simulations in refined temporal and spatial resolutions. Recent developments in measurement and sensor technologies have significantly improved the coverage, quality, frequency and diversity of hydrological observations. Data assimilation provides a great potential in relation to efficient use of traditional and new observational data in integrated hydrological models, as this technique can improve model prediction and reduce model uncertainty.The thesis investigates several challenges within the scope of data assimilation in integrated hydrological models. From the methodological point of view, different assimilation methodologies and techniques have been developed or customized to better serve hydrological assimilation. From the application point of view, real data and real-world complex catchments are used with the focus of investigating the models’ improvements with data assimilation.First, a systematic analysis was carried out for the case where groundwater hydraulic heads are assimilated in an integrated hydrological model. A data assimilation framework was developed and tested on synthetic data, and proved to be robust. It improved state estimation and was able to handle a variety of model uncertainty sources and scales. Next the groundwater head assimilation experiment was tested in a much more complex catchment with assimilation of biased real observations. In such cases, the bias-aware assimilation method significantly outperforms the standard assimilation method with improved estimation of state and observation bias. Multivariate assimilation of soil moisture and groundwater head was also investigated on the same two catchments representing simple and more complicated settings respectively. It was demonstrated that localization is able to reduce the impact of spurious correlation both within and between variables, and improves the results in both the unsaturated and saturated zones. Moreover, by combining multivariate assimilation with integrated hydrological models, the improvements can also be seen for other hydrological variables including river discharge and evapotranspiration. Overall, the study has documented a great potential for data assimilation in integrated hydrological modelling.

AB - Integrated hydrological models are frequently used in water-related environmental resource management. With our better understanding of the hydrological processes and the improved computational power, hydrological models are becoming increasingly more complex as they integrate multiple hydrological processes and provide simulations in refined temporal and spatial resolutions. Recent developments in measurement and sensor technologies have significantly improved the coverage, quality, frequency and diversity of hydrological observations. Data assimilation provides a great potential in relation to efficient use of traditional and new observational data in integrated hydrological models, as this technique can improve model prediction and reduce model uncertainty.The thesis investigates several challenges within the scope of data assimilation in integrated hydrological models. From the methodological point of view, different assimilation methodologies and techniques have been developed or customized to better serve hydrological assimilation. From the application point of view, real data and real-world complex catchments are used with the focus of investigating the models’ improvements with data assimilation.First, a systematic analysis was carried out for the case where groundwater hydraulic heads are assimilated in an integrated hydrological model. A data assimilation framework was developed and tested on synthetic data, and proved to be robust. It improved state estimation and was able to handle a variety of model uncertainty sources and scales. Next the groundwater head assimilation experiment was tested in a much more complex catchment with assimilation of biased real observations. In such cases, the bias-aware assimilation method significantly outperforms the standard assimilation method with improved estimation of state and observation bias. Multivariate assimilation of soil moisture and groundwater head was also investigated on the same two catchments representing simple and more complicated settings respectively. It was demonstrated that localization is able to reduce the impact of spurious correlation both within and between variables, and improves the results in both the unsaturated and saturated zones. Moreover, by combining multivariate assimilation with integrated hydrological models, the improvements can also be seen for other hydrological variables including river discharge and evapotranspiration. Overall, the study has documented a great potential for data assimilation in integrated hydrological modelling.

UR - https://soeg.kb.dk/permalink/45KBDK_KGL/fbp0ps/alma99122663412805763

M3 - Ph.D. thesis

BT - Data Assimilation in Integrated and Distributed Hydrological Models

PB - Department of Geosciences and Natural Resource Management, Faculty of Science, University of Copenhagen

ER -