The aim of this paper is to check the precision, accuracy and repeatability of network-based real time kinematic (NRTK) positioning under favorable (rover located inside the active network, close to the reference stations and at a similar altitude) and nonfavorable conditions (near to the boundaries region and at greater distances and elevation differences from the nearest reference stations). The rover location with respect to the reference stations is a crucial factor, especially in boundaries regions where favorable geometry cannot be assured. The elevation difference between the rover position and the nearest reference stations is also decisive. If the residual tropospheric delay is not modeled carefully there will be a bias error in the vertical component. This affects the precision of NRTK positioning, especially in the altimetric component. In order to analyze the NRTK positioning based on the Andalusian positioning network (RAP), three “nonsimulated” network scenarios (real outside and inside test sites/data are used) with different topography and environmental conditions in Southern Spain are presented: central area of Andalusia with full network coverage and the northern and southern borders of the region which are characterized as a mountainous and a coastal area, respectively. In the last two the network corrections are extrapolated. The statistical results confirm that it is possible to achieve centimeter-scale accuracy with NRTK positioning based on the RAP network even in border and coastal areas of Andalusia. However, the numerical results for each test site may be taken into consideration in proposing local improvements in the RAP network.