The use of distributed-physically based hydrological models is often hampered by the lack of information on key parameters and their spatial distribution and temporal dynamics. Typically, the estimation of parameter values is impeded by the lack of sufficient observations leading to mathematically underdetermined estimation problems and thus non-uniqueness. Catchment tomography (CT) presents a method to estimate spatially distributed model parameters by resolving the integrated signal of stream runoff in response to precipitation. Basically CT exploits the information content generated by a distributed precipitation signal both in time and space. In a moving transmitter-receiver concept, high resolution, radar based precipitation data are applied with a distributed surface runoff model. Synthetic stream water level observations, serving as receivers, are assimilated with an Ensemble Kalman Filter. With a joint state-parameter update the spatially distributed Manning's roughness coefficient, n, is estimated using the coupled Terrestrial Systems Modelling Platform and the Parallel Data Assimilation Framework (TerrSysMP-PDAF). The sequential data assimilation in combination with the distributed precipitation continuously integrates new information into the model, thus, increasingly constraining the parameter space. With this large amount of data included for the parameter estimation, CT reduces the problem of underdetermined model parameters. The initially biased Manning's coefficients spatially distributed in two and four fixed parameter zones are estimated with errors of less than 3% and 17%, respectively, with only 64 model realizations. It is shown that the distributed precipitation is of major importance for this approach.