Network-based Differential GPS (DGPS), regardless of its global, regional or local scales, is enabling technology to improve GPS positioning accuracy from tens of meters, to the levels of meters, decimetres and centimetres level in real time, depending on geographical coverage of the network and measurement types. The method is to use the data from a permanent network of reference stations to model errors due to inaccurate GPS satellite orbit ephemeris and clock data, ionospheric and tropospheric effects as well as other GPS satellite and receiver biases. Then error correction messages can be sent to users via any communication link in real time. This PhD research involves algorithm development for generating satellite orbit and tropospheric delay corrections using a regional or local reference network, especially tropospheric grid corrections, which have not been included in the existing DGPS correction vector messages, for the next generation of regional GNSS positioning services. Contributions of the research are made in the following three areas: First of all, research has been undertaken to test orbit interpolation methods, in order to represent GPS orbits and orbital corrections accurately and efficiently for (near) real-time GPS applications. For precise and predicted GPS orbits given in SP3 format and orbital corrections with respect to the broadcast ephemeris, numerical tests were conducted using different terms of Lagrange, Chebyshev and trigonometric polynomial functions. Secondly, this research has implemented a short-arc (9-hour) sliding-window orbit monitoring strategy to identify larger orbit errors in the predicted part of IGS ultra-rapid orbit solutions in near real time, using GPS tracking data from a regional network around Australia. The strategy is to predict the uncertainty estimates of each orbit over a short orbit arc in near real time, which allows users to down-weight the problematic satellites and reduces the effects of orbital errors for improved near real time ZTD estimation. Unlike long-arc orbit determination, we only estimate 6 orbital elements for each satellite. Finally, this research has proposed a new tropospheric delay correction model, which uses the Ordinary Kriging (OK) method to interpolate the residual ZTD within a regional area GPS network to improve the positioning accuracy. ZTD estimates from 129 EUREF Permanent Network (EPN) stations across Europe for over 3 months and from 17 GPSnet reference stations (Victoria, Australia) for one week were collected and processed for this study, respectively. It is concluded that interpolating residual ZTD is an efficient way to improve regional area differential GPS positioning.