Rainfall is one of the main inputs for hydrological models. Among the rainfall measuring sensors, raingauges and ground radars are probably the two most important in rainfall estimation. Therefore, the development of improved methodologies to estimate rainfall merging radar and raingauges data has been an objective from the initial studies of hydrological applications of meteorological radar. Thus, some previous works have reported that rainfall estimated by cokriging improved flood estimates (e.g. Sun, et. al., 2000). However, real time hydrological models have very restrictive requirements about maximum computing time that usually cokriging does not fulfill. Additionally the high variability on the rainfall field in time introduces additional problems in selecting and using a valid covariance (or variogram) model. All these problems make that cokriging have not been really applied on real time applications. The aim of this work is to provide an improved geostatistical approach able to be applied operationally. First, to avoid the traditional a priori selection of covariance models, we obtained valid covariance tables at each time step using an automatic nonparametric methodology based on FFT. Then, four alternatives are analyzed to test its efficiency in terms of time computing. The selected approaches used are Ordinary Kriging (OK), Kriging with External Drift (KED), CoKriging (COK), and Collocated CoKriging (ColCOK). 10-minute radar scans of a 22-hour rainfall case study in Catalonia is used to test the performances of these methodologies. For this event, the results show that rainfall estimated by ColCOK and KED give the best results, in a statistical and qualitative way, and that these methodologies can be used in a real time scheme application.