Tropospheric refraction is one of the major error sources in satellite-based positioning. The delay of radio signals caused by the troposphere ranges from 2 m at the zenith to 20 m at low elevation angles, depending on pressure, temperature and humidity along the path of the signal transmission. If the delay is not properly modelled, positioning accuracy can degrade significantly. Empirical tropospheric models, with or without meteorological observations, are used to correct these delays but they cannot model tropospheric variations exactly since they are limited in accuracy and spatial resolution resulting in up to a few decimetres error in positioning solutions. The present availability of dense ground based Global Navigation Satellite System (GNSS) networks and the state of the art GNSS processing techniques enable precise estimation of Zenith Tropospheric Delays (ZTD) with different latency ranging from Near Real-Time (NRT) to post-processing. We describe a method for computing ZTD correction fields interpolating, through Ordinary Kriging, the residuals between GNSS-derived and model-computed ZTD at continuously operating GNSS stations. At a known user location, the correction which is added to the modelled-ZTD value can be obtained through a bi-linear interpolation with the four nearest grid points surrounding it. The performance of the method has been evaluated over a 1-year period at 25 European stations belonging to the EUREF and IGS network. It is found that such an empirical tropospheric model can be improved when considering tropospheric corrections coming from ground based GNSS network.