To enable precise understanding and prediction of transport in porous media, physical, chemical and biological processes and the interaction between these processes have to be considered. This is an interdisciplinary problem. In the scope of this work the focus is on the physically mmpelling force of the transport, the groundwater flow velocity. A groundwater flow velocimeter (Fig. 1), using the point dilution technique, has been developed to measure in situ the groundwater flow velocity. The resolution of the groundwater flow velocimeter in the vertical direction is 25 cm. The in situ dilution measurement of the tracer uranin is carried out by laser-induced fluorimetry. The calibration of the groundwater flow velocimeter and subsequent field experiments at Krauthausen test site showed that the groundwater flow velocity inside a borehole can be measured with sufficient accuracy. The accuracy in deriving the Darcy velocity from the measurements inside a borehole is strongly dependent on the accuracy in determining the $\alpha$-factor, which corrects for the convergence of streamlines towards the borehole. The estimation of the $\alpha$-factor is difficult due to a general lack of knowledge of the state of the well. Measurements with the groundwater velocimeter showed that, the local Darcy velocity is strongly space and time dependent. For the strongly anisotropic structure of the hydraulic conductivity at the Krauthausen test site, statistics for a three dimensional heterogeneous flow velocity field have been estimated using stochastic theories and Monte Carlo analysis which is based on numerical calculations. It is shown that correlograms in the mean flow direction, estimated by 1st order approximation, are in agreement with the results from the Monte Carlo analysis. The variances for the Darcy velocity components, estimated by 1st order approximation, are clearly below the results of the Monte Carlo estimation. Estimation of the variance, based on 2nd order approximation, showed accordance only for the longitudinal component of the Darcy velocity, whereas the vertical transversal component is below and the horizontal transversal component is above the results from Monte Carlo analysis. The estimated mean Darcy velocity is similar for 1st and 2nd order approximation as well as for the Monte Carlo analysis. Estimation of the Darcy velocity statistics, both from numerical modelling and from stochastic theories, show smaller mean by factor 2, smaller variances by factor 4 and longer correlation length in the horizontal direction by factor 7 than the directly measured Darcy velocity using the groundwater flow velocimeter. The high discrepancies in mean, variance and horizontal autocorrelation length are due to the limited knowledge of the $\alpha$-factor and the time variability of the Darcy velocities. A comparison shows that the relative standard deviation is similar for modelling, 2nd order approximation and direct measurements with the groundwater flow velocimeter.