The availability of Earth observation (EO) and numerical weather prediction data for hydrological modelling and water management has increased significantly, creating a situation that today, for the same variable, estimates may be available from two or more sources of information. Precipitation data, for example, can be obtained from rain gauges, weather radar, satellites, or outputs from numerical weather models. Land use data can be obtained from land survey, satellite imagery, or a combination of the two. Each of these data sources provides an estimate of a catchment characteristic and related hydrological model parameters, or of a hydrometeorological variable. Estimates from each data source vary in magnitude or temporal and spatial variability. It is not always possible to judge which data source is the most accurate. One data source may perform poorly in one situation but give an accurate estimate for another. Yet, in hydrological modelling, usually, a particular set of catchment characteristics and input data is selected, possibly ignoring other relevant data sources. One of the reasons may be that despite vast research and development efforts in integration methods for sub-sets of the available data sources, there is no comprehensive data-model integration framework assuming existence and enabling effective use of multiple data sources in hydrological modelling. The main objective of this thesis, therefore, is to develop such a data-model integration framework, and test it on a case study.@en

Data for water management is increasingly easy to access, it has finer spatial and temporal resolution, and it is available from various sources. Precipitation data can be obtained from meteorological stations, radar, satellites and weather models. Land use data is also available from different satellite products and different providers. The various sources of data may confirm each other or give very different values in space and time. However, from these various data sources, it can often not be judged beforehand that one data is correct and others are wrong. Each source has its own value for a particular purpose. The Rijnland area in the Netherlands is one of the areas for which various data sources are available. Data sources that are researched in this paper are precipitation from rain gauges and radar, and three different land use maps. Various sources of data are used as input to the hydrological model (SIMGRO) of the water system to produce different discharge model output. Each run provides a member of the ensemble simulation which are combined to improve prediction of discharge from the catchment. It is shown that even simple averaging allows for increasing the model accuracy. @en