The increasing number of modernized GNSS signals and the availability of multi-constellation receivers are crucial for improvements of both precision and robustness of GNSS based positioning. However, the abundance of GNSS observations is not always useable as applications, using differential positioning or other techniques, may have limitations with respect to computational resources or communication bandwidth for reference data, and therefore require a qualified selection of a subset of observations for positioning. This paper is based on the work conducted in the project PREParE SHIPS funded by the European Union Agency for the Space Programme (EUSPA) on the specific application of Maritime Navigation using Network Real Time Kinematic (NRTK) and will focus on the satellite selection algorithms of the Prepare Ships dissemination solution. This study is motivated by data rate requirements and restrictions of the VDES dissemination solution developed in Prepare Ships. The restricted data rate for dissemination of RTK observations via VDES implies the need for a qualified pre-selection of satellite subsets to match the available bandwidth and the requirements of the positioning system. For this, multiple algorithms have been developed and tested in static and dynamic scenarios. Optimization techniques for height (for vertical position), two and three dimensions were examined. Different weighting schemes were used. During the evolution of the satellite selection study, it was concluded that it is necessary to retain satellites with the highest elevation as this will empirically improve integer ambiguity resolution for position fixing. Also fixing a minimum number of satellites for each constellation was required to enable a fair weightage to the different constellations used. Such algorithms should prove to be very useful for research on various Network RTK applications which require/prefer limited bandwidth such as for cadastral surveying and mapping, for airborne geo-referencing of aerial mapping data using Unmanned Aerial Vehicles (UAV) and on the road and sea for positioning and navigation of automated transport. Additionally, these algorithms could also be extended to consider satellite visibility in e.g. urban areas (i.e. urban canyons) by inclusion of true surface information for more robust GNSS positioning in automated transport applications . This could either be for pre-evaluation or for dynamically considering spatial information. While this work is a part of PREParE SHIPS, it is also motivated by a more general applicability of the algorithms presented for other similar applications. RTK correction dissemination with limited bandwidth requirements is very promising for RTK research and therefore this study on optimized selection of satellite subsets is of vital importance and could tap multiple opportunities of huge potential such as those involving NRTK or combination of Precise Point Positioning with RTK.